Adding upstream version 86.0.1.upstream/86.0.1upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

65 files changed, 5186 insertions, 0 deletions

diff --git a/js/src/gdb/README b/js/src/gdb/README
new file mode 100644
index 0000000000..9c98e192b5
--- /dev/null
+++ b/js/src/gdb/README
@@ -0,0 +1,198 @@
+This directory holds Python code to support debugging SpiderMonkey with
+GDB. It includes pretty-printers for common SpiderMonkey types like JS::Value,
+jsid, and JSObject, and makes GDB "see through" the SpiderMonkey rooting
+types like js::Rooted and JS::Handle. For example:
+
+    (gdb) frame
+    #0  js::baseops::SetPropertyHelper (cx=0xbf3460,
+        obj=(JSObject * const) 0x7ffff150b060 [object global] delegate,
+        receiver=(JSObject * const) 0x7ffff150b060 [object global] delegate,
+        id=$jsid("x"), defineHow=4, vp=$JS::Int32Value(1), strict=0)
+        at /home/jimb/moz/archer/js/src/jsobj.cpp:4495
+    4495	    MOZ_ASSERT((defineHow & ~(DNP_CACHE_RESULT | DNP_UNQUALIFIED)) == 0);
+    (gdb)
+
+Things to note here:
+
+- obj, a JS::HandleObject, prints as:
+      obj=(JSObject * const) 0x7ffff150b060 [object global] delegate,
+  This immediately shows the handle's referent, along with a JavaScript-like summary
+  of the object.
+
+- id, a JS::HandleId, prints as:
+      id=$jsid("x"),
+  We show the handle's referent, and print the identifier as a string.
+
+- vp, a JS::MutableHandleValue, prints as:
+      vp=$JS::Int32Value(1)
+  We show the handle's referent, using the JS::Value's tag to print it noting
+  its particular internal type and value.
+
+You can still see the raw form of a value with 'print/r':
+
+    (gdb) p/r obj
+    $1 = {<js::HandleBase<JSObject*>> = {<No data fields>}, ptr = 0x7fffffffca60}
+    (gdb)
+
+You can also use GDB's 'disable pretty-printer' command to turn off
+individual pretty-printers; try 'info pretty-printer' first.
+
+GDB should pick these extensions up automatically when you debug the shell or
+the browser, by auto-loading the 'js-gdb.py' file that the build system
+installs alongside the 'js' executable (or 'libxul.so-gdb.py' for the browser).
+You may need to add a command like the following to your '$HOME/.gdbinit' file:
+
+    # Tell GDB to trust auto-load files found under ~/moz.
+    add-auto-load-safe-path ~/moz
+
+If you do need this, GDB will tell you.
+
+In general, pretty-printers for pointer types include a summary of the
+pointer's referent:
+
+    (gdb) b math_atan2
+    Breakpoint 1 at 0x542e0a: file /home/jimb/moz/archer/js/src/jsmath.cpp, line 214.
+    (gdb) run
+    js> Math.atan2('Spleen', 42)
+    Breakpoint 1, math_atan2 (cx=0xbf3440, argc=2, vp=0x7ffff172f0a0)
+    (gdb) print vp[0]
+    $1 = $JS::Value((JSObject *) 0x7ffff151c0c0 [object Function "atan2"])
+    (gdb) print vp[1]
+    $2 = $JS::Value((JSObject *) 0x7ffff150d0a0 [object Math])
+    (gdb) print vp[2]
+    $3 = $JS::Value("Spleen")
+    (gdb) print vp[3]
+    $4 = $JS::Int32Value(42)
+    (gdb)
+
+We used to also have pretty-printers for the actual contents of a JSString
+struct, that knew which union branches were live and which were dead. These were
+more fragile than the summary pretty-printers, and harder to test, so I've
+removed them until we can see how to do better.
+
+There are unit tests; see 'Running the unit tests', below.
+
+I'd love for others to pitch in. GDB's Python API is documented in the GDB
+manual.
+
+I've recently rewritten the printers. The new code is simpler, and more
+robust; unit tests are easier to write; and the new test harness can run
+the tests in parallel. If a printer you'd contributed to in the past was
+dropped in the process, I apologize; I felt we should have good test
+coverage for any printer landed in-tree. You may also be interested in
+'Personal pretty-printers', below.
+
+Directory layout
+----------------
+
+- js/src/gdb/mozilla: The actual SpiderMonkey support code. GDB auto-loads this
+  when you debug an executable or shared library that contains SpiderMonkey.
+- js/src/gdb/tests: Unit tests for the above.
+  - Each '.py' file is a unit test, to be run by js/src/gdb/run-tests.py.
+  - Each '.cpp' file contains C++ code fragments for some unit test to use.
+- js/src/gdb/lib-for-tests: Python modules used by the unit tests.
+
+In js/src/gdb:
+
+- run-tests.py: test harness for GDB SpiderMonkey support unit tests. See
+  'Running the unit tests', below.
+- taskpool.py, progressbar.py: Python modules used by run-tests.py.
+- gdb-tests.cpp, gdb-tests.h: Driver program for C++ code fragments.
+- gdb-tests-gdb.py.in: Template for GDB autoload file for gdb-tests.
+
+Personal pretty-printers
+------------------------
+
+If you'd like to write your own pretty-printers, you can put them in a
+module named 'my_mozilla_printers' in a directory somewhere on your Python
+module search path. Our autoload code tries to import 'my_mozilla_printers'
+after importing our other SpiderMonkey support modules. For example:
+
+    $ echo $PYTHONPATH
+    /home/jimb/python
+    $ cat ~/python/my_mozilla_printers.py
+    import gdb
+    from mozilla.prettyprinters import ptr_pretty_printer
+
+    # Simple char16_t * printer. Doesn't show address; chases null pointers.
+    @ptr_pretty_printer('char16_t')
+    class char16Ptr(object):
+        def __init__(self, value, cache): self.value = value
+        def display_hint(self): return 'string'
+        def to_string(self):
+            c = u''
+            for i in range(50):
+                if self.value[i] == 0: break
+                c += unichr(self.value[i])
+            return c
+    $
+    ...
+    (gdb) whatis sample
+    type = char16_t [4]
+    (gdb) print &sample[0]
+    $1 = "Hi!"
+
+Running the unit tests
+----------------------
+
+These extensions have unit tests, invoked as follows:
+
+$ python run-tests.py [OPTIONS] OBJDIR [TESTS...]
+
+where OBJDIR is a directory containing a standalone SpiderMonkey build; TESTS
+are names of selected tests to run (if omitted, we run them all); and OPTIONS
+are drawn from the list below.
+
+--gdb=EXECUTABLE
+  Instead of running whatever 'gdb' we find in our search path, use
+  EXECUTABLE to run the tests.
+
+--srcdir=SRCDIR
+  Find the sources corresponding to OBJDIR/dist/bin/libmozjs.so in SRCDIR.
+  Without this option, we use the parent of the directory containing
+  'run-tests.py'. Note that SRCDIR must be a complete SpiderMonkey source
+  directory, as our tests #include internal SpiderMonkey header files (to
+  test pretty-printers for internal types, like parse nodes.)
+
+--testdir=TESTDIR
+  Search for Python scripts and any accompanying C++ source code in
+  TESTDIR. If omitted, we use the 'tests' directory in the directory
+  containing 'run-tests.py'.
+
+--builddir=BUILDDIR
+  Build the C++ executable that GDB debugs to run the tests in BUILDDIR.
+  If omitted, create a 'gdb-tests' subdirectory of OBJDIR/js/src.
+
+(It is safe to use relative paths for OBJDIR, SRCDIR, and so on. They are
+always interpreted relative to the directory that was current when
+run-tests.py was started.)
+
+For example, since I build in a subdirectory 'obj~' of the 'js/src'
+directory, I use this command from 'js/src' to run the pretty-printer unit
+tests:
+
+  $ python gdb/run-tests.py obj~
+
+Writing new unit tests
+----------------------
+
+Each unit test consists of a Python script, possibly with some accompanying
+C++ code. Running tests works like this:
+
+- The run-tests.py script calls 'make' in 'BUILDDIR/gdb' to build
+  'gdb-tests'.
+
+- Then, for each '.py' test script in js/src/gdb/tests, the harness starts
+  GDB on the 'gdb-tests' executable, and then has GDB run
+  js/src/gdb/lib-for-tests/prologue.py, passing it the test script's path as
+  its first command-line argument.
+
+Thanks To:
+----------
+
+- David Anderson
+- Steve Fink
+- Chris Leary
+- Josh Matthews
+- Jason Orendorff
+- Andrew Sutherland
diff --git a/js/src/gdb/TODO b/js/src/gdb/TODO
new file mode 100644
index 0000000000..07186aa108
--- /dev/null
+++ b/js/src/gdb/TODO
@@ -0,0 +1,22 @@
+* Ideas:
+- char16_t *
+- js::Shape, js::Baseshape
+- printers for structures with horrible unions (JSString, JSParseNode)
+- bring back parse_node.py
+- New 'js show' command for showing full trees, property lists, hash table
+  contents, and so on --- JSParseNode * should not show the whole tree.
+  Possibly clean up some "pointer-only" stuff in parse_node.py.
+  - 'js show <defn>' lists a JSDefinition's uses
+  - 'js show <parsenode>' shows entire tree
+  - 'js show <scope>' lists all properties (parents)
+  - 'js tree <scope>' shows property tree
+- avoid dead union branches in js::Shape; print attrs nicely
+- Print JSScope with identifier.
+- Print JSAtomSets, and thus PN_NAMESET.
+- JSParseNode PN_NAMESET
+- 'JSClass *' pretty-printer
+
+
+Local variables:
+mode: org
+End:
diff --git a/js/src/gdb/gdb-tests-gdb.py.in b/js/src/gdb/gdb-tests-gdb.py.in
new file mode 100644
index 0000000000..c6e4009816
--- /dev/null
+++ b/js/src/gdb/gdb-tests-gdb.py.in
@@ -0,0 +1,11 @@
+"""GDB Python customization auto-loader for GDB test executable"""
+#filter substitution
+
+import os.path
+sys.path[0:0] = [os.path.join('@topsrcdir@', 'gdb')]
+
+import mozilla.autoload
+mozilla.autoload.register(gdb.current_objfile())
+
+import mozilla.asmjs
+mozilla.asmjs.install()
diff --git a/js/src/gdb/gdb-tests.cpp b/js/src/gdb/gdb-tests.cpp
new file mode 100644
index 0000000000..641161582f
--- /dev/null
+++ b/js/src/gdb/gdb-tests.cpp
@@ -0,0 +1,93 @@
+/* This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "gdb-tests.h"
+#include "jsapi.h"
+#include "jsfriendapi.h"
+#include "js/Initialization.h"
+#include "js/Warnings.h"  // JS::SetWarningReporter
+
+using namespace JS;
+
+/* The class of the global object. */
+static const JSClass global_class = {"global", JSCLASS_GLOBAL_FLAGS,
+                                     &DefaultGlobalClassOps};
+
+static volatile int dontOptimizeMeAway = 0;
+
+void usePointer(const void* ptr) { dontOptimizeMeAway++; }
+
+template <typename T>
+static inline T* checkPtr(T* ptr) {
+  if (!ptr) {
+    abort();
+  }
+  return ptr;
+}
+
+static void checkBool(bool success) {
+  if (!success) {
+    abort();
+  }
+}
+
+/* The warning reporter callback. */
+void reportWarning(JSContext* cx, JSErrorReport* report) {
+  fprintf(stderr, "%s:%u: %s\n",
+          report->filename ? report->filename : "<no filename>",
+          (unsigned int)report->lineno, report->message().c_str());
+}
+
+// prologue.py sets a breakpoint on this function; test functions can call it
+// to easily return control to GDB where desired.
+void breakpoint() {
+  // If we leave this function empty, the linker will unify it with other
+  // empty functions throughout SpiderMonkey. If we then set a GDB
+  // breakpoint on it, that breakpoint will hit at all sorts of random
+  // times. So make it perform a distinctive side effect.
+  fprintf(stderr, "Called " __FILE__ ":breakpoint\n");
+}
+
+GDBFragment* GDBFragment::allFragments = nullptr;
+
+int main(int argc, const char** argv) {
+  if (!JS_Init()) return 1;
+  JSContext* cx = checkPtr(JS_NewContext(1024 * 1024));
+
+  JS_SetGCParameter(cx, JSGC_MAX_BYTES, 0xffffffff);
+  JS_SetNativeStackQuota(cx, 5000000);
+
+  checkBool(JS::InitSelfHostedCode(cx));
+  JS::SetWarningReporter(cx, reportWarning);
+
+  /* Create the global object. */
+  JS::RealmOptions options;
+  RootedObject global(
+      cx, checkPtr(JS_NewGlobalObject(cx, &global_class, nullptr,
+                                      JS::FireOnNewGlobalHook, options)));
+  JSAutoRealm ar(cx, global);
+
+  argv++;
+  while (*argv) {
+    const char* name = *argv++;
+    GDBFragment* fragment;
+    for (fragment = GDBFragment::allFragments; fragment;
+         fragment = fragment->next) {
+      if (strcmp(fragment->name(), name) == 0) {
+        fragment->run(cx, argv);
+        break;
+      }
+    }
+    if (!fragment) {
+      fprintf(stderr, "Unrecognized fragment name: %s\n", name);
+      exit(1);
+    }
+  }
+
+  return 0;
+}
diff --git a/js/src/gdb/gdb-tests.h b/js/src/gdb/gdb-tests.h
new file mode 100644
index 0000000000..7d32f8f949
--- /dev/null
+++ b/js/src/gdb/gdb-tests.h
@@ -0,0 +1,96 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
+ * vim: set ts=8 sts=2 et sw=2 tw=80:
+ *
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+#ifndef gdb_gdb_tests_h
+#define gdb_gdb_tests_h
+
+// Support for C++ fragments to be used by Python unit tests for SpiderMonkey's
+// GDB support.
+//
+// That is:
+// - js/src/gdb/mozilla holds the actual GDB SpiderMonkey support code.
+// - Each '.py' file in js/src/gdb/tests is a unit test for the above.
+// - Each '.cpp' file in js/src/gdb/tests is C++ code for one of the unit tests
+//   to run.
+//
+// (So the .cpp files are two steps removed from being anything one would
+// actually run.)
+
+#include "NamespaceImports.h"
+
+#include "js/GCAnnotations.h"
+
+void breakpoint();
+
+extern void usePointer(const void* ptr);
+
+template <typename T>
+void use(const T& thing) {
+  usePointer(&thing);
+}
+
+struct AutoSuppressHazardsForTest {
+  int dummy;
+  AutoSuppressHazardsForTest() : dummy(3) {}
+  ~AutoSuppressHazardsForTest() {
+    // Need nontrivial destructor.
+    usePointer(&dummy);
+  }
+} JS_HAZ_GC_SUPPRESSED;
+
+struct GDBFragment {
+  GDBFragment() {
+    next = allFragments;
+    allFragments = this;
+  }
+
+  // The name of this fragment. gdb-tests.cpp runs the fragments whose names
+  // are passed to it on the command line.
+  virtual const char* name() = 0;
+
+  // Run the fragment code. |argv| is a reference to the pointer into the
+  // command-line argument vector, referring to the argument immediately
+  // following this fragment's name. The fragment can consume arguments and
+  // advance argv if it wishes.
+  virtual void run(JSContext* cx, const char**& argv) = 0;
+
+  // We declare one instance of this type for each fragment to run. The
+  // constructor adds each instance to a linked list, of which this is
+  // the head.
+  static GDBFragment* allFragments;
+
+  // The link in the list of all instances.
+  GDBFragment* next;
+};
+
+// Macro for declaring a C++ fragment for some Python unit test to call. Usage:
+//
+//   FRAGMENT(<category>, <name>) { <body of fragment function> }
+//
+// where <category> and <name> are identifiers. The gdb-tests executable
+// takes a series of fragment names as command-line arguments and runs them in
+// turn; each fragment is named <category>.<name> on the command line.
+//
+// The body runs in a scope where 'cx' is a usable JSContext*.
+
+#define FRAGMENT(category, subname)                                   \
+  class FRAGMENT_CLASS_NAME(category, subname) : public GDBFragment { \
+    void run(JSContext* cx, const char**& argv) override;             \
+    const char* name() override {                                     \
+      return FRAGMENT_STRING_NAME(category, subname);                 \
+    }                                                                 \
+    static FRAGMENT_CLASS_NAME(category, subname) singleton;          \
+  };                                                                  \
+  FRAGMENT_CLASS_NAME(category, subname)                              \
+  FRAGMENT_CLASS_NAME(category, subname)::singleton;                  \
+  void FRAGMENT_CLASS_NAME(category, subname)::run(JSContext* cx,     \
+                                                   const char**& argv)
+
+#define FRAGMENT_STRING_NAME(category, subname) (#category "." #subname)
+#define FRAGMENT_CLASS_NAME(category, subname) Fragment_##category##_##subname
+
+#endif /* gdb_gdb_tests_h */
diff --git a/js/src/gdb/lib-for-tests/catcher.py b/js/src/gdb/lib-for-tests/catcher.py
new file mode 100644
index 0000000000..8ef3529255
--- /dev/null
+++ b/js/src/gdb/lib-for-tests/catcher.py
@@ -0,0 +1,35 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Apparently, there's simply no way to ask GDB to exit with a non-zero
+# status when the script run with the --eval-command option fails. Thus, if
+# we have --eval-command run prologue.py directly, syntax errors there will
+# lead GDB to exit with no indication anything went wrong.
+#
+# To avert that, we use this very small launcher script to run prologue.py
+# and catch errors.
+#
+# Remember, errors in this file will cause spurious passes, so keep this as
+# simple as possible!
+# flake8: noqa: F821
+
+import os
+import sys
+import traceback
+
+
+def execfile(filename, globs, locs):
+    with open(filename) as f:
+        code = compile(f.read(), filename, "exec")
+        exec(code, globs, locs)
+
+
+try:
+    # testlibdir is set on the GDB command line, via:
+    # --eval-command python testlibdir=...
+    execfile(os.path.join(testlibdir, "prologue.py"), globals(), locals())
+except Exception as err:
+    sys.stderr.write("Error running GDB prologue:\n")
+    traceback.print_exc()
+    sys.exit(1)
diff --git a/js/src/gdb/lib-for-tests/prologue.py b/js/src/gdb/lib-for-tests/prologue.py
new file mode 100644
index 0000000000..64f2994abe
--- /dev/null
+++ b/js/src/gdb/lib-for-tests/prologue.py
@@ -0,0 +1,128 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# flake8: noqa: F821
+
+import gdb
+import re
+import sys
+import traceback
+
+# testlibdir is set on the GDB command line, via --eval-command python testlibdir=...
+sys.path[0:0] = [testlibdir]
+
+active_fragment = None
+
+# Run the C++ fragment named |fragment|, stopping on entry to |function|
+# ('breakpoint', by default) and then select the calling frame.
+
+
+def run_fragment(fragment, function="gdb-tests.cpp:breakpoint"):
+    # Arrange to stop at a reasonable place in the test program.
+    bp = gdb.Breakpoint(function)
+    try:
+        gdb.execute("run %s" % (fragment,))
+        # Check that we did indeed stop by hitting the breakpoint we set.
+        assert bp.hit_count == 1
+    finally:
+        bp.delete()
+    gdb.execute("frame 1")
+
+    global active_fragment
+    active_fragment = fragment
+
+
+# Assert that |actual| is equal to |expected|; if not, complain in a helpful way.
+
+
+def assert_eq(actual, expected):
+    if actual != expected:
+        raise AssertionError(
+            """Unexpected result:
+expected: %r
+actual:   %r"""
+            % (expected, actual)
+        )
+
+
+# Assert that |expected| regex matches |actual| result; if not, complain in a helpful way.
+
+
+def assert_match(actual, expected):
+    if re.match(expected, actual, re.MULTILINE) is None:
+        raise AssertionError(
+            """Unexpected result:
+expected pattern: %r
+actual:           %r"""
+            % (expected, actual)
+        )
+
+
+# Assert that |value|'s pretty-printed form is |form|. If |value| is a
+# string, then evaluate it with gdb.parse_and_eval to produce a value.
+
+
+def assert_pretty(value, form):
+    if isinstance(value, str):
+        value = gdb.parse_and_eval(value)
+    assert_eq(str(value), form)
+
+
+# Assert that |value|'s pretty-printed form match the pattern |pattern|. If
+# |value| is a string, then evaluate it with gdb.parse_and_eval to produce a
+# value.
+
+
+def assert_regexp_pretty(value, form):
+    if isinstance(value, str):
+        value = gdb.parse_and_eval(value)
+    assert_match(str(value), form)
+
+
+# Check that the list of registered pretty-printers includes one named
+# |printer|, with a subprinter named |subprinter|.
+
+
+def assert_subprinter_registered(printer, subprinter):
+    # Match a line containing |printer| followed by a colon, and then a
+    # series of more-indented lines containing |subprinter|.
+
+    names = {"printer": re.escape(printer), "subprinter": re.escape(subprinter)}
+    pat = r"^( +)%(printer)s *\n(\1 +.*\n)*\1 +%(subprinter)s *\n" % names
+    output = gdb.execute("info pretty-printer", to_string=True)
+    if not re.search(pat, output, re.MULTILINE):
+        raise AssertionError(
+            "assert_subprinter_registered failed to find pretty-printer:\n"
+            "  %s:%s\n"
+            "'info pretty-printer' says:\n"
+            "%s" % (printer, subprinter, output)
+        )
+
+
+# Request full stack traces for Python errors.
+gdb.execute("set python print-stack full")
+
+# Tell GDB not to ask the user about the things we tell it to do.
+gdb.execute("set confirm off", False)
+
+# Some print settings that make testing easier.
+gdb.execute("set print static-members off")
+gdb.execute("set print address off")
+gdb.execute("set print pretty off")
+gdb.execute("set width 0")
+
+try:
+    # testscript is set on the GDB command line, via:
+    # --eval-command python testscript=...
+    execfile(testscript, globals(), locals())
+except AssertionError as err:
+    header = "\nAssertion traceback"
+    if active_fragment:
+        header += " for " + active_fragment
+    sys.stderr.write(header + ":\n")
+    (t, v, tb) = sys.exc_info()
+    traceback.print_tb(tb)
+    sys.stderr.write("\nTest assertion failed:\n")
+    sys.stderr.write(str(err))
+    sys.exit(1)
diff --git a/js/src/gdb/moz.build b/js/src/gdb/moz.build
new file mode 100644
index 0000000000..29c93a8fc0
--- /dev/null
+++ b/js/src/gdb/moz.build
@@ -0,0 +1,65 @@
+# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*-
+# vim: set filetype=python:
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this
+# file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+GeckoProgram("gdb-tests", linkage=None)
+
+include("../js-cxxflags.mozbuild")
+include("../js-standalone.mozbuild")
+
+SOURCES += [
+    # This file must not be unified with any of the test files, or the use()
+    # and breakpoint() functions might get optimized out.
+    "gdb-tests.cpp"
+]
+
+UNIFIED_SOURCES += [
+    "tests/enum-printers.cpp",
+    "tests/test-asmjs.cpp",
+    "tests/test-ExecutableAllocator.cpp",
+    "tests/test-GCCellPtr.cpp",
+    "tests/test-Interpreter.cpp",
+    "tests/test-jsbytecode.cpp",
+    "tests/test-jsid.cpp",
+    "tests/test-JSObject.cpp",
+    "tests/test-jsop.cpp",
+    "tests/test-JSString.cpp",
+    "tests/test-JSSymbol.cpp",
+    "tests/test-jsval.cpp",
+    "tests/test-Root.cpp",
+    "tests/test-unwind.cpp",
+    "tests/typedef-printers.cpp",
+]
+
+SOURCES += ["tests/test-prettyprinters.cpp"]
+
+if CONFIG["CC_TYPE"] != "clang-cl":
+    # Test expects to see pre-typedef names of base classes, but the compiler will
+    # normally omit those from the debuginfo. The current clang-cl does not support
+    # this option.
+    SOURCES["tests/test-prettyprinters.cpp"].flags += [
+        "-fno-eliminate-unused-debug-types"
+    ]
+
+if CONFIG["CC_TYPE"] == "clang":
+    # clang has poor debuginfo when optimized. Some of the test files have
+    # unrecoverable values even with -Og. gcc is far better about making
+    # optimized code debuggable.
+    CXXFLAGS += ["-O0"]
+
+DEFINES["EXPORT_JS_API"] = True
+
+LOCAL_INCLUDES += [
+    "!..",
+    "..",
+]
+
+USE_LIBS += [
+    "static:js",
+]
+
+DEFINES["topsrcdir"] = "%s/js/src" % TOPSRCDIR
+FINAL_TARGET_PP_FILES += ["gdb-tests-gdb.py.in"]
+OBJDIR_FILES.js.src.gdb += ["!/dist/bin/gdb-tests-gdb.py"]
diff --git a/js/src/gdb/mozilla/CellHeader.py b/js/src/gdb/mozilla/CellHeader.py
new file mode 100644
index 0000000000..8163e57e1b
--- /dev/null
+++ b/js/src/gdb/mozilla/CellHeader.py
@@ -0,0 +1,25 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+import gdb
+
+
+def get_header_ptr(value, ptr_t):
+    # Return the pointer stored in Cell::header_ for subclasses of
+    # TenuredCellWithNonGCPointer and CellWithTenuredGCPointer.
+    return value["header_"]["mValue"].cast(ptr_t)
+
+
+def get_header_length_and_flags(value, cache):
+    # Return the length and flags values for subclasses of
+    # CellWithLengthAndFlags.
+    flags = value["header_"]["mValue"].cast(cache.uintptr_t)
+    try:
+        length = value["length_"]
+    except gdb.error:
+        # If we couldn't fetch the length directly, it must be stored
+        # within `flags`.
+        length = flags >> 32
+        flags = flags % 2 ** 32
+    return length, flags
diff --git a/js/src/gdb/mozilla/ExecutableAllocator.py b/js/src/gdb/mozilla/ExecutableAllocator.py
new file mode 100644
index 0000000000..e82b1a186d
--- /dev/null
+++ b/js/src/gdb/mozilla/ExecutableAllocator.py
@@ -0,0 +1,96 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+"""
+All jitted code is allocated via the ExecutableAllocator class. Make GDB aware
+of them, such that we can query for pages which are containing code which are
+allocated by the Jits.
+"""
+
+import gdb
+import mozilla.prettyprinters
+from mozilla.prettyprinters import pretty_printer, ptr_pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+
+class jsjitExecutableAllocatorCache(object):
+    """Cache information about the ExecutableAllocator type for this objfile."""
+
+    def __init__(self):
+        self.d = None
+
+    def __getattr__(self, name):
+        if self.d is None:
+            self.initialize()
+        return self.d[name]
+
+    def initialize(self):
+        self.d = {}
+        self.d["ExecutableAllocator"] = gdb.lookup_type("js::jit::ExecutableAllocator")
+        self.d["ExecutablePool"] = gdb.lookup_type("js::jit::ExecutablePool")
+        self.d["HashNumber"] = gdb.lookup_type("mozilla::HashNumber")
+
+
+@pretty_printer("js::jit::ExecutableAllocator")
+class jsjitExecutableAllocator(object):
+    def __init__(self, value, cache):
+        if not cache.mod_ExecutableAllocator:
+            cache.mod_ExecutableAllocator = jsjitExecutableAllocatorCache()
+        self.value = value
+        self.cache = cache.mod_ExecutableAllocator
+
+    def to_string(self):
+        return "ExecutableAllocator([%s])" % ", ".join([str(x) for x in self])
+
+    def __iter__(self):
+        return self.PoolIterator(self)
+
+    class PoolIterator(object):
+        def __init__(self, allocator):
+            self.allocator = allocator
+            self.entryType = allocator.cache.ExecutablePool.pointer()
+            self.hashNumType = allocator.cache.HashNumber
+            # Emulate the HashSet::Range
+            self.table = allocator.value["m_pools"]["mImpl"]["mTable"]
+            self.index = 0
+            kHashNumberBits = 32
+            hashShift = allocator.value["m_pools"]["mImpl"]["mHashShift"]
+            self.capacity = 1 << (kHashNumberBits - hashShift)
+            if self.table == 0:
+                self.capacity = 0
+            # auto hashes = reinterpret_cast<HashNumber*>(mTable);
+            self.hashes = self.table.cast(self.hashNumType.pointer())
+            # auto entries = reinterpret_cast<Entry*>(&hashes[capacity()]);
+            self.entries = (self.hashes + self.capacity).cast(self.entryType.pointer())
+
+        def __iter__(self):
+            return self
+
+        def next(self):
+            return self.__next__()
+
+        def __next__(self):
+            cur = self.index
+            if cur >= self.capacity:
+                raise StopIteration()
+            self.index = self.index + 1
+            if self.hashes[cur] > 1:  # table[i]->isLive()
+                return self.entries[cur]
+            return self.__next__()
+
+
+@ptr_pretty_printer("js::jit::ExecutablePool")
+class jsjitExecutablePool(mozilla.prettyprinters.Pointer):
+    def __init__(self, value, cache):
+        if not cache.mod_ExecutableAllocator:
+            cache.mod_ExecutableAllocator = jsjitExecutableAllocatorCache()
+        self.value = value
+        self.cache = cache.mod_ExecutableAllocator
+
+    def to_string(self):
+        pages = self.value["m_allocation"]["pages"]
+        size = self.value["m_allocation"]["size"]
+        return "ExecutablePool %08x-%08x" % (pages, pages + size)
diff --git a/js/src/gdb/mozilla/GCCellPtr.py b/js/src/gdb/mozilla/GCCellPtr.py
new file mode 100644
index 0000000000..738d9b3e45
--- /dev/null
+++ b/js/src/gdb/mozilla/GCCellPtr.py
@@ -0,0 +1,121 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for GCCellPtr values.
+
+import gdb
+import mozilla.prettyprinters
+
+from mozilla.prettyprinters import pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+# Cache information about the types for this objfile.
+
+
+class GCCellPtrTypeCache(object):
+    def __init__(self, cache):
+        self.TraceKind_t = gdb.lookup_type("JS::TraceKind")
+        self.AllocKind_t = gdb.lookup_type("js::gc::AllocKind")
+        self.Arena_t = gdb.lookup_type("js::gc::Arena")
+        self.Cell_t = gdb.lookup_type("js::gc::Cell")
+        self.TenuredCell_t = gdb.lookup_type("js::gc::TenuredCell")
+
+        trace_kinds = gdb.types.make_enum_dict(self.TraceKind_t)
+        alloc_kinds = gdb.types.make_enum_dict(self.AllocKind_t)
+
+        def trace_kind(k):
+            return trace_kinds["JS::TraceKind::" + k]
+
+        def alloc_kind(k):
+            return alloc_kinds["js::gc::AllocKind::" + k]
+
+        # Build a mapping from TraceKind enum values to the types they denote.
+        trace_map = {
+            # Inline types.
+            "Object": "JSObject",
+            "BigInt": "JS::BigInt",
+            "String": "JSString",
+            "Symbol": "JS::Symbol",
+            "Shape": "js::Shape",
+            "ObjectGroup": "js::ObjectGroup",
+            "Null": "std::nullptr_t",
+            # Out-of-line types.
+            "BaseShape": "js::BaseShape",
+            "JitCode": "js::jit::JitCode",
+            "Script": "js::BaseScript",
+            "Scope": "js::Scope",
+            "RegExpShared": "js::RegExpShared",
+        }
+
+        # Map from AllocKind to TraceKind for out-of-line types.
+        alloc_map = {
+            "BASE_SHAPE": "BaseShape",
+            "JITCODE": "JitCode",
+            "SCRIPT": "Script",
+            "SCOPE": "Scope",
+            "REGEXP_SHARED": "RegExpShared",
+        }
+
+        self.trace_kind_to_type = {
+            trace_kind(k): gdb.lookup_type(v) for k, v in trace_map.items()
+        }
+        self.alloc_kind_to_trace_kind = {
+            alloc_kind(k): trace_kind(v) for k, v in alloc_map.items()
+        }
+
+        self.Null = trace_kind("Null")
+        self.tracekind_mask = gdb.parse_and_eval("JS::OutOfLineTraceKindMask")
+        self.arena_mask = gdb.parse_and_eval("js::gc::ArenaMask")
+
+
+@pretty_printer("JS::GCCellPtr")
+class GCCellPtr(object):
+    def __init__(self, value, cache):
+        self.value = value
+        if not cache.mod_GCCellPtr:
+            cache.mod_GCCellPtr = GCCellPtrTypeCache(cache)
+        self.cache = cache
+
+    def to_string(self):
+        ptr = self.value["ptr"]
+        kind = ptr & self.cache.mod_GCCellPtr.tracekind_mask
+        if kind == self.cache.mod_GCCellPtr.Null:
+            return "JS::GCCellPtr(nullptr)"
+        if kind == self.cache.mod_GCCellPtr.tracekind_mask:
+            # Out-of-line trace kinds.
+            #
+            # Compute the underlying type for out-of-line kinds by
+            # reimplementing the GCCellPtr::outOfLineKind() method.
+            #
+            # The extra casts below are only present to make it easier to
+            # compare this code against the C++ implementation.
+
+            # GCCellPtr::asCell()
+            cell_ptr = ptr & ~self.cache.mod_GCCellPtr.tracekind_mask
+            cell = cell_ptr.reinterpret_cast(self.cache.mod_GCCellPtr.Cell_t.pointer())
+
+            # Cell::asTenured()
+            tenured = cell.cast(self.cache.mod_GCCellPtr.TenuredCell_t.pointer())
+
+            # TenuredCell::arena()
+            addr = int(tenured)
+            arena_ptr = addr & ~self.cache.mod_GCCellPtr.arena_mask
+            arena = arena_ptr.reinterpret_cast(
+                self.cache.mod_GCCellPtr.Arena_t.pointer()
+            )
+
+            # Arena::getAllocKind()
+            alloc_kind = arena["allocKind"].cast(self.cache.mod_GCCellPtr.AllocKind_t)
+            alloc_idx = int(
+                alloc_kind.cast(self.cache.mod_GCCellPtr.AllocKind_t.target())
+            )
+
+            # Map the AllocKind to a TraceKind.
+            kind = self.cache.mod_GCCellPtr.alloc_kind_to_trace_kind[alloc_idx]
+        type_name = self.cache.mod_GCCellPtr.trace_kind_to_type[int(kind)]
+        return "JS::GCCellPtr(({}*) {})".format(
+            type_name, ptr.cast(self.cache.void_ptr_t)
+        )
diff --git a/js/src/gdb/mozilla/Interpreter.py b/js/src/gdb/mozilla/Interpreter.py
new file mode 100644
index 0000000000..febf92b86d
--- /dev/null
+++ b/js/src/gdb/mozilla/Interpreter.py
@@ -0,0 +1,93 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for InterpreterRegs.
+
+import gdb
+import mozilla.prettyprinters as prettyprinters
+
+prettyprinters.clear_module_printers(__name__)
+
+from mozilla.prettyprinters import pretty_printer
+
+
+class InterpreterTypeCache(object):
+    # Cache information about the Interpreter types for this objfile.
+    def __init__(self):
+        self.tValue = gdb.lookup_type("JS::Value")
+        self.tJSOp = gdb.lookup_type("JSOp")
+        try:
+            self.tScriptFrameIterData = gdb.lookup_type("js::ScriptFrameIter::Data")
+        except gdb.error:
+            # Work around problem with gcc optimized debuginfo where it doesn't
+            # seem to be able to see that ScriptFrameIter inherits the
+            # FrameIter::Data type.
+            self.tScriptFrameIterData = gdb.lookup_type("js::FrameIter::Data")
+        self.tInterpreterFrame = gdb.lookup_type("js::InterpreterFrame")
+        self.tBaselineFrame = gdb.lookup_type("js::jit::BaselineFrame")
+        self.tRematerializedFrame = gdb.lookup_type("js::jit::RematerializedFrame")
+        self.tDebugFrame = gdb.lookup_type("js::wasm::DebugFrame")
+
+
+@pretty_printer("js::InterpreterRegs")
+class InterpreterRegs(object):
+    def __init__(self, value, cache):
+        self.value = value
+        self.cache = cache
+        if not cache.mod_Interpreter:
+            cache.mod_Interpreter = InterpreterTypeCache()
+        self.itc = cache.mod_Interpreter
+
+    # There's basically no way to co-operate with 'set print pretty' (how would
+    # you get the current level of indentation?), so we don't even bother
+    # trying. No 'children', just 'to_string'.
+    def to_string(self):
+        fp_ = "fp_ = {}".format(self.value["fp_"])
+        slots = (self.value["fp_"] + 1).cast(self.itc.tValue.pointer())
+        sp = "sp = fp_.slots() + {}".format(self.value["sp"] - slots)
+        pc = "pc = {}".format(self.value["pc"])
+        return "{{ {}, {}, {} }}".format(fp_, sp, pc)
+
+
+@pretty_printer("js::AbstractFramePtr")
+class AbstractFramePtr(object):
+    Tag_ScriptFrameIterData = 0x0
+    Tag_InterpreterFrame = 0x1
+    Tag_BaselineFrame = 0x2
+    Tag_RematerializedFrame = 0x3
+    Tag_WasmDebugFrame = 0x4
+    TagMask = 0x7
+
+    def __init__(self, value, cache):
+        self.value = value
+        self.cache = cache
+        if not cache.mod_Interpreter:
+            cache.mod_Interpreter = InterpreterTypeCache()
+        self.itc = cache.mod_Interpreter
+
+    def to_string(self):
+        ptr = self.value["ptr_"]
+        tag = ptr & AbstractFramePtr.TagMask
+        ptr = ptr & ~AbstractFramePtr.TagMask
+        if tag == AbstractFramePtr.Tag_ScriptFrameIterData:
+            label = "js::ScriptFrameIter::Data"
+            ptr = ptr.cast(self.itc.tScriptFrameIterData.pointer())
+        if tag == AbstractFramePtr.Tag_InterpreterFrame:
+            label = "js::InterpreterFrame"
+            ptr = ptr.cast(self.itc.tInterpreterFrame.pointer())
+        if tag == AbstractFramePtr.Tag_BaselineFrame:
+            label = "js::jit::BaselineFrame"
+            ptr = ptr.cast(self.itc.tBaselineFrame.pointer())
+        if tag == AbstractFramePtr.Tag_RematerializedFrame:
+            label = "js::jit::RematerializedFrame"
+            ptr = ptr.cast(self.itc.tRematerializedFrame.pointer())
+        if tag == AbstractFramePtr.Tag_WasmDebugFrame:
+            label = "js::wasm::DebugFrame"
+            ptr = ptr.cast(self.itc.tDebugFrame.pointer())
+        return "AbstractFramePtr (({} *) {})".format(label, ptr)
+
+    # Provide the ptr_ field as a child, so it prints after the pretty string
+    # provided above.
+    def children(self):
+        yield ("ptr_", self.value["ptr_"])
diff --git a/js/src/gdb/mozilla/IonGraph.py b/js/src/gdb/mozilla/IonGraph.py
new file mode 100644
index 0000000000..b5f6a899b8
--- /dev/null
+++ b/js/src/gdb/mozilla/IonGraph.py
@@ -0,0 +1,281 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+"""
+Debugging JIT Compilations can be obscure without large context. This python
+script provide commands to let GDB open an image viewer to display the graph of
+any compilation, as they are executed within GDB.
+
+This python script should be sourced within GDB after loading the python scripts
+provided with SpiderMonkey.
+"""
+
+import gdb
+import os
+import subprocess
+import tempfile
+import time
+import mozilla.prettyprinters
+from mozilla.prettyprinters import pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+# Cache information about the JSString type for this objfile.
+
+
+class jsvmPrinterCache(object):
+    def __init__(self):
+        self.d = None
+
+    def __getattr__(self, name):
+        if self.d is None:
+            self.initialize()
+        return self.d[name]
+
+    def initialize(self):
+        self.d = {}
+        self.d["char"] = gdb.lookup_type("char")
+
+
+# Dummy class used to store the content of the type cache in the context of the
+# iongraph command, which uses the jsvmLSprinter.
+
+
+class ModuleCache(object):
+    def __init__(self):
+        self.mod_IonGraph = None
+
+
+@pretty_printer("js::vm::LSprinter")
+class jsvmLSprinter(object):
+    def __init__(self, value, cache):
+        self.value = value
+        if not cache.mod_IonGraph:
+            cache.mod_IonGraph = jsvmPrinterCache()
+        self.cache = cache.mod_IonGraph
+
+    def to_string(self):
+        next = self.value["head_"]
+        tail = self.value["tail_"]
+        if next == 0:
+            return ""
+        res = ""
+        while next != tail:
+            chars = (next + 1).cast(self.cache.char.pointer())
+            res = res + chars.string("ascii", "ignore", next["length"])
+            next = next["next"]
+        length = next["length"] - self.value["unused_"]
+        chars = (next + 1).cast(self.cache.char.pointer())
+        res = res + chars.string("ascii", "ignore", length)
+        return res
+
+
+def search_in_path(bin):
+    paths = os.getenv("PATH", "")
+    if paths == "":
+        return ""
+    for d in paths.split(":"):
+        f = os.path.join(d, bin)
+        if os.access(f, os.X_OK):
+            return f
+    return ""
+
+
+class IonGraphBinParameter(gdb.Parameter):
+    set_doc = "Set the path to iongraph binary, used by iongraph command."
+    show_doc = "Show the path to iongraph binary, used by iongraph command."
+
+    def get_set_string(self):
+        return "Path to iongraph binary changed to: %s" % self.value
+
+    def get_show_string(self, value):
+        return "Path to iongraph binary set to: %s" % value
+
+    def __init__(self):
+        super(IonGraphBinParameter, self).__init__(
+            "iongraph-bin", gdb.COMMAND_SUPPORT, gdb.PARAM_FILENAME
+        )
+        self.value = os.getenv("GDB_IONGRAPH", "")
+        if self.value == "":
+            self.value = search_in_path("iongraph")
+
+
+class DotBinParameter(gdb.Parameter):
+    set_doc = "Set the path to dot binary, used by iongraph command."
+    show_doc = "Show the path to dot binary, used by iongraph command."
+
+    def get_set_string(self):
+        return "Path to dot binary changed to: %s" % self.value
+
+    def get_show_string(self, value):
+        return "Path to dot binary set to: %s" % value
+
+    def __init__(self):
+        super(DotBinParameter, self).__init__(
+            "dot-bin", gdb.COMMAND_SUPPORT, gdb.PARAM_FILENAME
+        )
+        self.value = os.getenv("GDB_DOT", "")
+        if self.value == "":
+            self.value = search_in_path("dot")
+
+
+class PngViewerBinParameter(gdb.Parameter):
+    set_doc = "Set the path to a png viewer binary, used by iongraph command."
+    show_doc = "Show the path to a png viewer binary, used by iongraph command."
+
+    def get_set_string(self):
+        return "Path to a png viewer binary changed to: %s" % self.value
+
+    def get_show_string(self):
+        return "Path to a png viewer binary set to: %s" % self.value
+
+    def __init__(self):
+        super(PngViewerBinParameter, self).__init__(
+            "pngviewer-bin", gdb.COMMAND_SUPPORT, gdb.PARAM_FILENAME
+        )
+        self.value = os.getenv("GDB_PNGVIEWER", "")
+        if self.value == "":
+            self.value = search_in_path("xdg-open")
+
+
+iongraph = IonGraphBinParameter()
+dot = DotBinParameter()
+pngviewer = PngViewerBinParameter()
+
+
+class IonGraphCommand(gdb.Command):
+    """Command used to display the current state of the MIR graph in a png
+    viewer by providing an expression to access the MIRGenerator.
+    """
+
+    def __init__(self):
+        super(IonGraphCommand, self).__init__(
+            "iongraph", gdb.COMMAND_DATA, gdb.COMPLETE_EXPRESSION
+        )
+        self.typeCache = ModuleCache()
+
+    def invoke(self, mirGenExpr, from_tty):
+        """Call function from the graph spewer to populate the json printer with
+        the content generated by the jsonSpewer. Then we read the json content
+        from the jsonPrinter internal data, and gives that as input of iongraph
+        command."""
+
+        # From the MIRGenerator, find the graph spewer which contains both the
+        # jsonPrinter (containing the result of the output), and the jsonSpewer
+        # (continaining methods for spewing the graph).
+        mirGen = gdb.parse_and_eval(mirGenExpr)
+        jsonPrinter = mirGen["gs_"]["jsonPrinter_"]
+        jsonSpewer = mirGen["gs_"]["jsonSpewer_"]
+        graph = mirGen["graph_"]
+
+        # These commands are doing side-effects which are saving the state of
+        # the compiled code on the LSprinter dedicated for logging. Fortunately,
+        # if you are using these gdb command, this probably means that other
+        # ways of getting this content failed you already, so making a mess in
+        # these logging strings should not cause much issues.
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).clear()"
+            % (
+                jsonPrinter.type,
+                jsonPrinter.address,
+            )
+        )
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).beginFunction((JSScript*)0)"
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+            )
+        )
+        gdb.parse_and_eval(
+            '(*(%s*)(%s)).beginPass("gdb")'
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+            )
+        )
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).spewMIR((%s)%s)"
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+                graph.type,
+                graph,
+            )
+        )
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).spewLIR((%s)%s)"
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+                graph.type,
+                graph,
+            )
+        )
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).endPass()"
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+            )
+        )
+        gdb.parse_and_eval(
+            "(*(%s*)(%s)).endFunction()"
+            % (
+                jsonSpewer.type,
+                jsonSpewer.address,
+            )
+        )
+
+        # Dump the content of the LSprinter containing the JSON view of the
+        # graph into a python string.
+        json = jsvmLSprinter(jsonPrinter, self.typeCache).to_string()
+
+        # We are in the middle of the program execution and are messing up with
+        # the state of the logging data. As this might not be the first time we
+        # call beginFunction, we might have an extra comma at the beginning of
+        # the output, just strip it.
+        if json[0] == ",":
+            json = json[1:]
+
+        # Usually this is added by the IonSpewer.
+        json = '{ "functions": [%s] }' % json
+
+        # Display the content of the json with iongraph and other tools.
+        self.displayMIRGraph(json)
+
+    def displayMIRGraph(self, jsonStr):
+        png = tempfile.NamedTemporaryFile()
+
+        # start all processes in a shell-like equivalent of:
+        #   iongraph < json | dot > tmp.png; xdg-open tmp.png
+        i = subprocess.Popen(
+            [iongraph.value, "--funcnum", "0", "--passnum", "0", "--out-mir", "-", "-"],
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+        )
+        d = subprocess.Popen([dot.value, "-Tpng"], stdin=i.stdout, stdout=png)
+
+        # Write the json file as the input of the iongraph command.
+        i.stdin.write(jsonStr.encode("utf8"))
+        i.stdin.close()
+        i.stdout.close()
+
+        # Wait for iongraph and dot, such that the png file contains all the
+        # bits needed to by the png viewer.
+        i.wait()
+        d.communicate()[0]
+
+        # Spawn & detach the png viewer, to which we give the name of the
+        # temporary file.  Note, as we do not want to wait on the image viewer,
+        # there is a minor race between the removal of the temporary file, which
+        # would happen at the next garbage collection cycle, and the start of
+        # the png viewer.  We could use a pipe, but unfortunately, this does not
+        # seems to be supported by xdg-open.
+        subprocess.Popen([pngviewer.value, png.name], stdin=None, stdout=None)
+        time.sleep(1)
+
+
+iongraph_cmd = IonGraphCommand()
diff --git a/js/src/gdb/mozilla/JSObject.py b/js/src/gdb/mozilla/JSObject.py
new file mode 100644
index 0000000000..282b47a5d1
--- /dev/null
+++ b/js/src/gdb/mozilla/JSObject.py
@@ -0,0 +1,86 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for SpiderMonkey JSObjects.
+
+import re
+import gdb
+import mozilla.prettyprinters as prettyprinters
+from mozilla.prettyprinters import ptr_pretty_printer, ref_pretty_printer
+from mozilla.Root import deref
+from mozilla.CellHeader import get_header_ptr
+
+prettyprinters.clear_module_printers(__name__)
+
+
+class JSObjectTypeCache(object):
+    def __init__(self, value, cache):
+        baseshape_flags = gdb.lookup_type("js::BaseShape::Flag")
+        self.flag_DELEGATE = prettyprinters.enum_value(
+            baseshape_flags, "js::BaseShape::DELEGATE"
+        )
+        self.func_ptr_type = gdb.lookup_type("JSFunction").pointer()
+        self.class_NON_NATIVE = gdb.parse_and_eval("JSClass::NON_NATIVE")
+        self.NativeObject_ptr_t = gdb.lookup_type("js::NativeObject").pointer()
+        self.BaseShape_ptr_t = gdb.lookup_type("js::BaseShape").pointer()
+        self.Shape_ptr_t = gdb.lookup_type("js::Shape").pointer()
+        self.ObjectGroup_ptr_t = gdb.lookup_type("js::ObjectGroup").pointer()
+        self.JSClass_ptr_t = gdb.lookup_type("JSClass").pointer()
+
+
+# There should be no need to register this for JSFunction as well, since we
+# search for pretty-printers under the names of base classes, and
+# JSFunction has JSObject as a base class.
+
+
+gdb_string_regexp = re.compile(r'(?:0x[0-9a-z]+ )?(?:<.*> )?"(.*)"', re.I)
+
+
+@ptr_pretty_printer("JSObject")
+class JSObjectPtrOrRef(prettyprinters.Pointer):
+    def __init__(self, value, cache):
+        super(JSObjectPtrOrRef, self).__init__(value, cache)
+        if not cache.mod_JSObject:
+            cache.mod_JSObject = JSObjectTypeCache(value, cache)
+        self.otc = cache.mod_JSObject
+
+    def summary(self):
+        group = get_header_ptr(self.value, self.otc.ObjectGroup_ptr_t)
+        classp = get_header_ptr(group, self.otc.JSClass_ptr_t)
+        non_native = classp["flags"] & self.otc.class_NON_NATIVE
+
+        # Use GDB to format the class name, but then strip off the address
+        # and the quotes.
+        class_name = str(classp["name"])
+        m = gdb_string_regexp.match(class_name)
+        if m:
+            class_name = m.group(1)
+
+        if non_native:
+            return "[object {}]".format(class_name)
+        else:
+            native = self.value.cast(self.otc.NativeObject_ptr_t)
+            shape = deref(native["shape_"])
+            baseshape = get_header_ptr(shape, self.otc.BaseShape_ptr_t)
+            flags = baseshape["flags"]
+            is_delegate = bool(flags & self.otc.flag_DELEGATE)
+            name = None
+            if class_name == "Function":
+                function = self.value
+                concrete_type = function.type.strip_typedefs()
+                if concrete_type.code == gdb.TYPE_CODE_REF:
+                    function = function.address
+                function = function.cast(self.otc.func_ptr_type)
+                atom = deref(function["atom_"])
+                name = str(atom) if atom else "<unnamed>"
+            return "[object {}{}]{}".format(
+                class_name,
+                " " + name if name else "",
+                " delegate" if is_delegate else "",
+            )
+
+
+@ref_pretty_printer("JSObject")
+def JSObjectRef(value, cache):
+    return JSObjectPtrOrRef(value, cache)
diff --git a/js/src/gdb/mozilla/JSString.py b/js/src/gdb/mozilla/JSString.py
new file mode 100644
index 0000000000..121480b1f8
--- /dev/null
+++ b/js/src/gdb/mozilla/JSString.py
@@ -0,0 +1,104 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for SpiderMonkey strings.
+
+import gdb
+import mozilla.prettyprinters
+from mozilla.prettyprinters import ptr_pretty_printer
+from mozilla.CellHeader import get_header_length_and_flags
+
+try:
+    chr(10000)  # UPPER RIGHT PENCIL
+except ValueError:  # yuck, we are in Python 2.x, so chr() is 8-bit
+    chr = unichr  # replace with teh unicodes
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+
+class JSStringTypeCache(object):
+    # Cache information about the JSString type for this objfile.
+    def __init__(self, cache):
+        dummy = gdb.Value(0).cast(cache.JSString_ptr_t)
+        self.ATOM_BIT = dummy["ATOM_BIT"]
+        self.LINEAR_BIT = dummy["LINEAR_BIT"]
+        self.INLINE_CHARS_BIT = dummy["INLINE_CHARS_BIT"]
+        self.TYPE_FLAGS_MASK = dummy["TYPE_FLAGS_MASK"]
+        self.LATIN1_CHARS_BIT = dummy["LATIN1_CHARS_BIT"]
+
+
+class Common(mozilla.prettyprinters.Pointer):
+    def __init__(self, value, cache):
+        super(Common, self).__init__(value, cache)
+        if not cache.mod_JSString:
+            cache.mod_JSString = JSStringTypeCache(cache)
+        self.stc = cache.mod_JSString
+
+
+@ptr_pretty_printer("JSString")
+class JSStringPtr(Common):
+    def display_hint(self):
+        return "string"
+
+    def chars(self):
+        d = self.value["d"]
+        length, flags = get_header_length_and_flags(self.value, self.cache)
+
+        corrupt = {
+            0x2F2F2F2F: "JS_FRESH_NURSERY_PATTERN",
+            0x2B2B2B2B: "JS_SWEPT_NURSERY_PATTERN",
+            0xE5E5E5E5: "jemalloc freed memory",
+        }.get(flags & 0xFFFFFFFF)
+        if corrupt:
+            for ch in "<CORRUPT:%s>" % corrupt:
+                yield ch
+            return
+        is_rope = (flags & self.stc.LINEAR_BIT) == 0
+        if is_rope:
+            for c in JSStringPtr(d["s"]["u2"]["left"], self.cache).chars():
+                yield c
+            for c in JSStringPtr(d["s"]["u3"]["right"], self.cache).chars():
+                yield c
+        else:
+            is_inline = (flags & self.stc.INLINE_CHARS_BIT) != 0
+            is_latin1 = (flags & self.stc.LATIN1_CHARS_BIT) != 0
+            if is_inline:
+                if is_latin1:
+                    chars = d["inlineStorageLatin1"]
+                else:
+                    chars = d["inlineStorageTwoByte"]
+            else:
+                if is_latin1:
+                    chars = d["s"]["u2"]["nonInlineCharsLatin1"]
+                else:
+                    chars = d["s"]["u2"]["nonInlineCharsTwoByte"]
+            for i in range(int(length)):
+                yield chars[i]
+
+    def to_string(self, maxlen=200):
+        s = ""
+        invalid_chars_allowed = 2
+        for c in self.chars():
+            if len(s) >= maxlen:
+                s += "..."
+                break
+
+            try:
+                # Convert from gdb.Value to string.
+                s += chr(c)
+            except ValueError:
+                if invalid_chars_allowed == 0:
+                    s += "<TOO_MANY_INVALID_CHARS>"
+                    break
+                else:
+                    invalid_chars_allowed -= 1
+                    s += "\\x%04x" % (c & 0xFFFF)
+        return s
+
+
+@ptr_pretty_printer("JSAtom")
+class JSAtomPtr(Common):
+    def to_string(self):
+        return self.value.cast(self.cache.JSString_ptr_t)
diff --git a/js/src/gdb/mozilla/JSSymbol.py b/js/src/gdb/mozilla/JSSymbol.py
new file mode 100644
index 0000000000..06b354404d
--- /dev/null
+++ b/js/src/gdb/mozilla/JSSymbol.py
@@ -0,0 +1,40 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printer for SpiderMonkey symbols.
+
+import mozilla.prettyprinters
+from mozilla.prettyprinters import ptr_pretty_printer
+from mozilla.CellHeader import get_header_ptr
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+# JS::SymbolCode enumerators
+PrivateNameSymbol = 0xFFFFFFFD
+InSymbolRegistry = 0xFFFFFFFE
+UniqueSymbol = 0xFFFFFFFF
+
+
+@ptr_pretty_printer("JS::Symbol")
+class JSSymbolPtr(mozilla.prettyprinters.Pointer):
+    def __init__(self, value, cache):
+        super(JSSymbolPtr, self).__init__(value, cache)
+        self.value = value
+
+    def to_string(self):
+        code = int(self.value["code_"]) & 0xFFFFFFFF
+        desc = str(get_header_ptr(self.value, self.cache.JSString_ptr_t))
+        if code == InSymbolRegistry:
+            return "Symbol.for({})".format(desc)
+        elif code == UniqueSymbol:
+            return "Symbol({})".format(desc)
+        elif code == PrivateNameSymbol:
+            return "#{}".format(desc)
+        else:
+            # Well-known symbol. Strip off the quotes added by the JSString *
+            # pretty-printer.
+            assert desc[0] == '"'
+            assert desc[-1] == '"'
+            return desc[1:-1]
diff --git a/js/src/gdb/mozilla/PropertyKey.py b/js/src/gdb/mozilla/PropertyKey.py
new file mode 100644
index 0000000000..5e7dc9c811
--- /dev/null
+++ b/js/src/gdb/mozilla/PropertyKey.py
@@ -0,0 +1,66 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for JSID values.
+
+import mozilla.prettyprinters
+import mozilla.Root
+
+from mozilla.prettyprinters import pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+
+@pretty_printer("JS::PropertyKey")
+class PropertyKey(object):
+    # Since people don't always build with macro debugging info, I can't
+    # think of any way to avoid copying these values here, short of using
+    # inferior calls for every operation (which, I hear, is broken from
+    # pretty-printers in some recent GDBs).
+    TYPE_STRING = 0x0
+    TYPE_INT = 0x1
+    TYPE_VOID = 0x2
+    TYPE_SYMBOL = 0x4
+    TYPE_EMPTY = 0x6
+    TYPE_MASK = 0x7
+
+    def __init__(self, value, cache):
+        self.value = value
+        self.cache = cache
+        self.concrete_type = self.value.type.strip_typedefs()
+
+    def to_string(self):
+        bits = self.value["asBits"]
+        tag = bits & PropertyKey.TYPE_MASK
+        if tag == PropertyKey.TYPE_STRING:
+            body = bits.cast(self.cache.JSString_ptr_t)
+        elif tag & PropertyKey.TYPE_INT:
+            body = bits >> 1
+        elif tag == PropertyKey.TYPE_VOID:
+            return "JSID_VOID"
+        elif tag == PropertyKey.TYPE_SYMBOL:
+            body = (bits & ~PropertyKey.TYPE_MASK).cast(self.cache.JSSymbol_ptr_t)
+        elif tag == PropertyKey.TYPE_EMPTY:
+            return "JSID_EMPTY"
+        else:
+            body = "<unrecognized>"
+        return "$jsid(%s)" % (body,)
+
+
+@pretty_printer("JS::Rooted<long>")
+def RootedPropertyKey(value, cache):
+    # Hard-code the referent type pretty-printer for PropertyKey roots and
+    # handles. See the comment for mozilla.Root.Common.__init__.
+    return mozilla.Root.Rooted(value, cache, PropertyKey)
+
+
+@pretty_printer("JS::Handle<long>")
+def HandlePropertyKey(value, cache):
+    return mozilla.Root.Handle(value, cache, PropertyKey)
+
+
+@pretty_printer("JS::MutableHandle<long>")
+def MutableHandlePropertyKey(value, cache):
+    return mozilla.Root.MutableHandle(value, cache, PropertyKey)
diff --git a/js/src/gdb/mozilla/Root.py b/js/src/gdb/mozilla/Root.py
new file mode 100644
index 0000000000..fa80f3fec6
--- /dev/null
+++ b/js/src/gdb/mozilla/Root.py
@@ -0,0 +1,104 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers and utilities for SpiderMonkey rooting templates:
+# Rooted, Handle, MutableHandle, etc.
+
+import mozilla.prettyprinters
+from mozilla.prettyprinters import template_pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+
+class Common(object):
+    # Common base class for all the rooting template pretty-printers. All these
+    # templates have one member holding the referent (or a pointer to it), so
+    # there's not much to it.
+
+    # The name of the template member holding the referent.
+    member = "ptr"
+
+    # If True, this is a handle type, and should be dereferenced. If False,
+    # the template member holds the referent directly.
+    handle = False
+
+    # If True, we should strip typedefs from our referent type. (Rooted<T>
+    # uses template magic that gives the referent a noisy type.)
+    strip_typedefs = False
+
+    # Initialize a pretty-printer for |value|, using |cache|.
+    #
+    # If given, |content_printer| is a pretty-printer constructor to use for
+    # this handle/root/etc.'s referent. Usually, we can just omit this argument
+    # and let GDB choose a pretty-printer for the referent given its type, but
+    # when the referent is a typedef of an integral type (say, |jsid| in a
+    # non-|DEBUG| build), the GNU toolchain (at least) loses the typedef name,
+    # and all we know about the referent is its fundamental integer type ---
+    # |JS::Rooted<jsid>|, for example, appears in GDB as |JS::Rooted<long>| ---
+    # and we are left with no way to choose a meaningful pretty-printer based on
+    # the type of the referent alone. However, because we know that the only
+    # integer type for which |JS::Rooted| is likely to be instantiated is
+    # |jsid|, we *can* register a pretty-printer constructor for the full
+    # instantiation |JS::Rooted<long>|. That constructor creates a |JS::Rooted|
+    # pretty-printer, and explicitly specifies the constructor for the referent,
+    # using this initializer's |content_printer| argument.
+    def __init__(self, value, cache, content_printer=None):
+        self.value = value
+        self.cache = cache
+        self.content_printer = content_printer
+
+    def to_string(self):
+        ptr = self.value[self.member]
+        if self.handle:
+            ptr = ptr.dereference()
+        if self.strip_typedefs:
+            ptr = ptr.cast(ptr.type.strip_typedefs())
+        if self.content_printer:
+            return self.content_printer(ptr, self.cache).to_string()
+        else:
+            # As of 2012-11, GDB suppresses printing pointers in replacement
+            # values; see http://sourceware.org/ml/gdb/2012-11/msg00055.html
+            # That means that simply returning the 'ptr' member won't work.
+            # Instead, just invoke GDB's formatter ourselves.
+            return str(ptr)
+
+
+@template_pretty_printer("JS::Rooted")
+class Rooted(Common):
+    strip_typedefs = True
+
+
+@template_pretty_printer("JS::Handle")
+class Handle(Common):
+    handle = True
+
+
+@template_pretty_printer("JS::MutableHandle")
+class MutableHandle(Common):
+    handle = True
+
+
+@template_pretty_printer("js::BarrieredBase")
+class BarrieredBase(Common):
+    member = "value"
+
+
+def deref(root):
+    # Return the referent of a HeapPtr, Rooted, or Handle.
+    tag = root.type.strip_typedefs().tag
+    if not tag:
+        raise TypeError(
+            "Can't dereference type with no structure tag: %s" % (root.type,)
+        )
+    elif tag.startswith("js::HeapPtr<"):
+        return root["value"]
+    elif tag.startswith("JS::Rooted<"):
+        return root["ptr"]
+    elif tag.startswith("JS::Handle<"):
+        return root["ptr"]
+    elif tag.startswith("js::GCPtr<"):
+        return root["value"]
+    else:
+        raise NotImplementedError("Unrecognized tag: " + tag)
diff --git a/js/src/gdb/mozilla/__init__.py b/js/src/gdb/mozilla/__init__.py
new file mode 100644
index 0000000000..7084d5052e
--- /dev/null
+++ b/js/src/gdb/mozilla/__init__.py
@@ -0,0 +1,5 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Yes, Python, this is a package.
diff --git a/js/src/gdb/mozilla/asmjs.py b/js/src/gdb/mozilla/asmjs.py
new file mode 100644
index 0000000000..85a19a5cfc
--- /dev/null
+++ b/js/src/gdb/mozilla/asmjs.py
@@ -0,0 +1,51 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+"""
+In asm code, out-of-bounds heap accesses cause segfaults, which the engine
+handles internally. Make GDB ignore them.
+"""
+
+import gdb
+
+SIGSEGV = 11
+
+# A sigaction buffer for each inferior process.
+sigaction_buffers = {}
+
+
+def on_stop(event):
+    if isinstance(event, gdb.SignalEvent) and event.stop_signal == "SIGSEGV":
+        # Allocate memory for sigaction, once per js shell process.
+        process = gdb.selected_inferior()
+        buf = sigaction_buffers.get(process)
+        if buf is None:
+            buf = gdb.parse_and_eval(
+                "(struct sigaction *) malloc(sizeof(struct sigaction))"
+            )
+            sigaction_buffers[process] = buf
+
+        # See if WasmFaultHandler is installed as the SIGSEGV signal action.
+        sigaction_fn = gdb.parse_and_eval(
+            "(void(*)(int,void*,void*))__sigaction"
+        ).dereference()
+        sigaction_fn(SIGSEGV, 0, buf)
+        WasmTrapHandler = gdb.parse_and_eval("WasmTrapHandler")
+        if buf["__sigaction_handler"]["sa_handler"] == WasmTrapHandler:
+            # Advise the user that magic is happening.
+            print("js/src/gdb/mozilla/asmjs.py: Allowing WasmTrapHandler to run.")
+
+            # If WasmTrapHandler doesn't handle this segfault, it will unhook
+            # itself and re-raise.
+            gdb.execute("continue")
+
+
+def on_exited(event):
+    if event.inferior in sigaction_buffers:
+        del sigaction_buffers[event.inferior]
+
+
+def install():
+    gdb.events.stop.connect(on_stop)
+    gdb.events.exited.connect(on_exited)
diff --git a/js/src/gdb/mozilla/autoload.py b/js/src/gdb/mozilla/autoload.py
new file mode 100644
index 0000000000..23f248762d
--- /dev/null
+++ b/js/src/gdb/mozilla/autoload.py
@@ -0,0 +1,40 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# mozilla/autoload.py: Autoload SpiderMonkey pretty-printers.
+
+print("Loading JavaScript value pretty-printers; see js/src/gdb/README.")
+print("If they cause trouble, type: disable pretty-printer .* SpiderMonkey")
+
+import gdb.printing
+import mozilla.prettyprinters
+
+# Import the pretty-printer modules. As a side effect, loading these
+# modules registers their printers with mozilla.prettyprinters.
+import mozilla.GCCellPtr
+import mozilla.ExecutableAllocator
+import mozilla.Interpreter
+import mozilla.IonGraph
+import mozilla.JSObject
+import mozilla.JSString
+import mozilla.JSSymbol
+import mozilla.Root
+import mozilla.PropertyKey
+import mozilla.jsop
+import mozilla.jsval
+import mozilla.unwind
+
+# The user may have personal pretty-printers. Get those, too, if they exist.
+try:
+    import my_mozilla_printers  # NOQA: F401
+except ImportError:
+    pass
+
+
+def register(objfile):
+    # Register our pretty-printers with |objfile|.
+    lookup = mozilla.prettyprinters.lookup_for_objfile(objfile)
+    if lookup:
+        gdb.printing.register_pretty_printer(objfile, lookup, replace=True)
+    mozilla.unwind.register_unwinder(objfile)
diff --git a/js/src/gdb/mozilla/jitsrc.py b/js/src/gdb/mozilla/jitsrc.py
new file mode 100644
index 0000000000..98e2b96db4
--- /dev/null
+++ b/js/src/gdb/mozilla/jitsrc.py
@@ -0,0 +1,146 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# This is a gdb extension to automate the process of tracing backwards in rr
+# from a jit instruction to the code that generated that instruction.
+#
+# Usage:
+#  (rr) x/i $pc
+#  => 0x240e954ac13a:	pushq  (%rbx)
+#  (rr) jitsrc 0x240e954ac13a
+
+import gdb
+import re
+
+# (base_name, hops, func_name, source_var, dest_var) tuples, such that :
+#   - `base_name`: a regex matching the name of the function that implements
+#       the actual write
+#   - `hops`: the number of stack frames between `base_name` and `func_name`
+#   - `func_name`: a regex matching the name of the function that calls memcpy
+#   - `source_var`: an expression that can be evaluated in the frame
+#       corresponding to `func_name` to get the source of the memcpy
+#   - `dest_var`: an expression that can be evaluated in the frame
+#       corresponding to `func_name` to get the destination of the memcpy
+#
+# If an invocation of `jitsrc` stops in the middle of a memcpy, the solution
+# is normally to add a new pattern here.
+patterns = [
+    (
+        "__memmove_avx_unaligned_erms",
+        1,
+        "js::jit::X86Encoding::BaseAssembler::executableCopy",
+        "src",
+        "dst",
+    ),
+    (
+        "__memcpy_avx_unaligned",
+        1,
+        "js::jit::X86Encoding::BaseAssembler::executableCopy",
+        "src",
+        "dst",
+    ),
+    ("__memmove_avx_unaligned_erms", 1, "arena_t::RallocSmallOrLarge", "aPtr", "ret"),
+    ("__memcpy_avx_unaligned", 1, "arena_t::RallocSmallOrLarge", "aPtr", "ret"),
+    (
+        "mozilla::detail::VectorImpl<.*>::new_<.*>",
+        3,
+        "mozilla::Vector<.*>::convertToHeapStorage",
+        "beginNoCheck()",
+        "newBuf",
+    ),
+    (
+        "__memmove_avx_unaligned_erms",
+        1,
+        "js::jit::AssemblerBufferWithConstantPools",
+        "&cur->instructions[0]",
+        "dest",
+    ),
+    (
+        "__memcpy_sse2_unaligned",
+        1,
+        "js::jit::AssemblerBufferWithConstantPools",
+        "&cur->instructions[0]",
+        "dest",
+    ),
+    (
+        "__memcpy_sse2_unaligned",
+        2,
+        "js::jit::AssemblerX86Shared::executableCopy",
+        "masm.m_formatter.m_buffer.m_buffer.mBegin",
+        "buffer",
+    ),
+    ("__memcpy_sse2_unaligned", 1, "arena_t::RallocSmallOrLarge", "aPtr", "ret"),
+    ("js::jit::X86Encoding::SetInt32", 0, "js::jit::X86Encoding::SetInt32", "0", "0"),
+    (
+        "js::jit::X86Encoding::SetPointer",
+        0,
+        "js::jit::X86Encoding::SetPointer",
+        "0",
+        "0",
+    ),
+    (
+        "<unnamed>",
+        1,
+        "js::jit::AssemblerBufferWithConstantPools<.*>::executableCopy",
+        "&cur->instructions[0]",
+        "dest",
+    ),
+]
+
+
+class JitSource(gdb.Command):
+    def __init__(self):
+        super(JitSource, self).__init__("jitsrc", gdb.COMMAND_RUNNING)
+        self.dont_repeat()
+
+    def disable_breakpoints(self):
+        self.disabled_breakpoints = [b for b in gdb.breakpoints() if b.enabled]
+        for b in self.disabled_breakpoints:
+            b.enabled = False
+
+    def enable_breakpoints(self):
+        for b in self.disabled_breakpoints:
+            b.enabled = True
+
+    def search_stack(self, base_name, hops, name, src, dst, address):
+        current_frame_name = gdb.newest_frame().name() or "<unnamed>"
+        if not re.match(base_name, current_frame_name):
+            return None
+        f = gdb.newest_frame()
+        for _ in range(hops):
+            f = f.older()
+        if not re.match(name, f.name()):
+            return None
+        f.select()
+        src_val = gdb.parse_and_eval(src)
+        dst_val = gdb.parse_and_eval(dst)
+        return hex(src_val + int(address, 16) - dst_val)
+
+    def next_address(self, old):
+        for pattern in patterns:
+            found = self.search_stack(*pattern, old)
+            if found:
+                return found
+        return None
+
+    def runback(self, address):
+        b = gdb.Breakpoint(
+            "*" + address, type=gdb.BP_WATCHPOINT, wp_class=gdb.WP_WRITE, internal=True
+        )
+        while b.hit_count == 0:
+            gdb.execute("rc", to_string=True)
+        b.delete()
+
+    def invoke(self, arg, from_tty):
+        args = gdb.string_to_argv(arg)
+        address = args[0]
+        self.disable_breakpoints()
+        while address:
+            self.runback(address)
+            address = self.next_address(address)
+        self.enable_breakpoints()
+
+
+# Register to the gdb runtime
+JitSource()
diff --git a/js/src/gdb/mozilla/jsop.py b/js/src/gdb/mozilla/jsop.py
new file mode 100644
index 0000000000..ae21aad5f6
--- /dev/null
+++ b/js/src/gdb/mozilla/jsop.py
@@ -0,0 +1,59 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for JSOp and jsbytecode values.
+
+import gdb
+import gdb.types
+import mozilla.prettyprinters
+from mozilla.prettyprinters import pretty_printer, ptr_pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+
+class JSOpTypeCache(object):
+    # Cache information about the JSOp type for this objfile.
+    def __init__(self, cache):
+        self.tJSOp = gdb.lookup_type("JSOp")
+
+    @classmethod
+    def get_or_create(cls, cache):
+        if not cache.mod_JSOp:
+            cache.mod_JSOp = cls(cache)
+        return cache.mod_JSOp
+
+
+@pretty_printer("JSOp")
+class JSOp(object):
+    def __init__(self, value, cache):
+        self.value = value
+        self.cache = cache
+        self.jotc = JSOpTypeCache.get_or_create(cache)
+
+    def to_string(self):
+        # JSOp's storage type is |uint8_t|, but gdb uses a signed value.
+        # Manually convert it to an unsigned value.
+        #
+        # https://sourceware.org/bugzilla/show_bug.cgi?id=25325
+        idx = int(self.value.cast(self.jotc.tJSOp.target()))
+        assert 0 <= idx and idx <= 255
+        fields = self.jotc.tJSOp.fields()
+        if idx < len(fields):
+            return fields[idx].name
+        return "(JSOp) {:d}".format(idx)
+
+
+@ptr_pretty_printer("jsbytecode")
+class JSBytecodePtr(mozilla.prettyprinters.Pointer):
+    def __init__(self, value, cache):
+        super(JSBytecodePtr, self).__init__(value, cache)
+        self.jotc = JSOpTypeCache.get_or_create(cache)
+
+    def to_string(self):
+        try:
+            opcode = str(self.value.dereference().cast(self.jotc.tJSOp))
+        except Exception:
+            opcode = "bad pc"
+        return "{} ({})".format(self.value.cast(self.cache.void_ptr_t), opcode)
diff --git a/js/src/gdb/mozilla/jsval.py b/js/src/gdb/mozilla/jsval.py
new file mode 100644
index 0000000000..022e59b363
--- /dev/null
+++ b/js/src/gdb/mozilla/jsval.py
@@ -0,0 +1,232 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Pretty-printers for SpiderMonkey's JS::Value.
+
+import gdb
+import gdb.types
+import struct
+import mozilla.prettyprinters
+from mozilla.prettyprinters import pretty_printer
+
+# Forget any printers from previous loads of this module.
+mozilla.prettyprinters.clear_module_printers(__name__)
+
+# Summary of the JS::Value type:
+#
+# JS::Value is a 64-bit discriminated union, with JSString*, JSObject*, IEEE
+# 64-bit floating-point, and 32-bit integer branches (and a few others).
+# JS::Value is 64 bits long on all architectures.
+#
+# The ECMAScript standard specifies that ECMAScript numbers are IEEE 64-bit
+# floating-point values. A JS::Value can represent any JavaScript number
+# value directly, without referring to additional storage, or represent an
+# object, string, or other ECMAScript value, and remember which type it is.
+# This may seem surprising: how can a 64-bit type hold all the 64-bit IEEE
+# values, and still distinguish them from objects, strings, and so on,
+# which have 64-bit addresses?
+#
+# This is possible for two reasons:
+#
+# - First, ECMAScript implementations aren't required to distinguish all
+#   the values the IEEE 64-bit format can represent. The IEEE format
+#   specifies many bitstrings representing NaN values, while ECMAScript
+#   requires only a single NaN value. This means we can use one IEEE NaN to
+#   represent ECMAScript's NaN, and use all the other IEEE NaNs to
+#   represent the other ECMAScript values.
+#
+#   (IEEE says that any floating-point value whose 11-bit exponent field is
+#   0x7ff (all ones) and whose 52-bit fraction field is non-zero is a NaN.
+#   So as long as we ensure the fraction field is non-zero, and save a NaN
+#   for ECMAScript, we have 2^52 values to play with.)
+#
+# - Second, on the only 64-bit architecture we support, x86_64, only the
+#   lower 48 bits of an address are significant. The upper sixteen bits are
+#   required to be the sign-extension of bit 48. Furthermore, user code
+#   always runs in "positive addresses": those in which bit 48 is zero. So
+#   we only actually need 47 bits to store all possible object or string
+#   addresses, even on 64-bit platforms.
+#
+# With a 52-bit fraction field, and 47 bits needed for the 'payload', we
+# have up to five bits left to store a 'tag' value, to indicate which
+# branch of our discriminated union is live.
+#
+# Thus, we define JS::Value representations in terms of the IEEE 64-bit
+# floating-point format:
+#
+# - Any bitstring that IEEE calls a number or an infinity represents that
+#   ECMAScript number.
+#
+# - Any bitstring that IEEE calls a NaN represents either an ECMAScript NaN
+#   or a non-number ECMAScript value, as determined by a tag field stored
+#   towards the most significant end of the fraction field (exactly where
+#   depends on the address size). If the tag field indicates that this
+#   JS::Value is an object, the fraction field's least significant end
+#   holds the address of a JSObject; if a string, the address of a
+#   JSString; and so on.
+#
+# On x86_64 only the lower 48 bits of an address are significant, and only
+# those values whose top bit is zero are used for user-space addresses. Thus
+# x86_64 addresses are effectively 47 bits long and fit nicely in the available
+# portion of the fraction field.
+#
+# See Value.h for full details.
+
+
+class Box(object):
+    def __init__(self, asBits, jtc):
+        self.asBits = asBits
+        self.jtc = jtc
+        # Value::asBits is uint64_t, but somehow the sign bit can be botched
+        # here, even though Python integers are arbitrary precision.
+        if self.asBits < 0:
+            self.asBits = self.asBits + (1 << 64)
+
+    # Return this value's type tag.
+    def tag(self):
+        raise NotImplementedError
+
+    # Return this value as a 32-bit integer, double, or address.
+    def as_uint32(self):
+        raise NotImplementedError
+
+    def as_double(self):
+        packed = struct.pack("q", self.asBits)
+        (unpacked,) = struct.unpack("d", packed)
+        return unpacked
+
+    def as_address(self):
+        raise NotImplementedError
+
+
+class Punbox(Box):
+    # Packed non-number boxing --- the format used on x86_64. It would be nice to
+    # simply call Value::toInt32, etc. here, but the debugger is likely to see many
+    # Values, and doing several inferior calls for each one seems like a bad idea.
+
+    FULL_WIDTH = 64
+    TAG_SHIFT = 47
+    PAYLOAD_MASK = (1 << TAG_SHIFT) - 1
+    TAG_MASK = (1 << (FULL_WIDTH - TAG_SHIFT)) - 1
+    TAG_MAX_DOUBLE = 0x1FFF0
+    TAG_TYPE_MASK = 0x0000F
+
+    def tag(self):
+        tag = self.asBits >> Punbox.TAG_SHIFT
+        if tag <= Punbox.TAG_MAX_DOUBLE:
+            return self.jtc.DOUBLE
+        else:
+            return tag & Punbox.TAG_TYPE_MASK
+
+    def as_uint32(self):
+        return int(self.asBits & ((1 << 32) - 1))
+
+    def as_address(self):
+        return gdb.Value(self.asBits & Punbox.PAYLOAD_MASK)
+
+
+class Nunbox(Box):
+    TAG_SHIFT = 32
+    TAG_CLEAR = 0xFFFF0000
+    PAYLOAD_MASK = 0xFFFFFFFF
+    TAG_TYPE_MASK = 0x0000000F
+
+    def tag(self):
+        tag = self.asBits >> Nunbox.TAG_SHIFT
+        if tag < Nunbox.TAG_CLEAR:
+            return self.jtc.DOUBLE
+        return tag & Nunbox.TAG_TYPE_MASK
+
+    def as_uint32(self):
+        return int(self.asBits & Nunbox.PAYLOAD_MASK)
+
+    def as_address(self):
+        return gdb.Value(self.asBits & Nunbox.PAYLOAD_MASK)
+
+
+class JSValueTypeCache(object):
+    # Cache information about the Value type for this objfile.
+
+    def __init__(self, cache):
+        # Capture the tag values.
+        d = gdb.types.make_enum_dict(gdb.lookup_type("JSValueType"))
+
+        # The enum keys are prefixed when building with some compilers (clang at
+        # a minimum), so use a helper function to handle either key format.
+        def get(key):
+            val = d.get(key)
+            if val is not None:
+                return val
+            return d["JSValueType::" + key]
+
+        self.DOUBLE = get("JSVAL_TYPE_DOUBLE")
+        self.INT32 = get("JSVAL_TYPE_INT32")
+        self.UNDEFINED = get("JSVAL_TYPE_UNDEFINED")
+        self.BOOLEAN = get("JSVAL_TYPE_BOOLEAN")
+        self.MAGIC = get("JSVAL_TYPE_MAGIC")
+        self.STRING = get("JSVAL_TYPE_STRING")
+        self.SYMBOL = get("JSVAL_TYPE_SYMBOL")
+        self.BIGINT = get("JSVAL_TYPE_BIGINT")
+        self.NULL = get("JSVAL_TYPE_NULL")
+        self.OBJECT = get("JSVAL_TYPE_OBJECT")
+
+        # Let self.magic_names be an array whose i'th element is the name of
+        # the i'th magic value.
+        d = gdb.types.make_enum_dict(gdb.lookup_type("JSWhyMagic"))
+        self.magic_names = list(range(max(d.values()) + 1))
+        for (k, v) in d.items():
+            self.magic_names[v] = k
+
+        # Choose an unboxing scheme for this architecture.
+        self.boxer = Punbox if cache.void_ptr_t.sizeof == 8 else Nunbox
+
+
+@pretty_printer("JS::Value")
+class JSValue(object):
+    def __init__(self, value, cache):
+        # Save the generic typecache, and create our own, if we haven't already.
+        self.cache = cache
+        if not cache.mod_JS_Value:
+            cache.mod_JS_Value = JSValueTypeCache(cache)
+        self.jtc = cache.mod_JS_Value
+
+        self.value = value
+        self.box = self.jtc.boxer(value["asBits_"], self.jtc)
+
+    def to_string(self):
+        tag = self.box.tag()
+
+        if tag == self.jtc.UNDEFINED:
+            return "$JS::UndefinedValue()"
+        if tag == self.jtc.NULL:
+            return "$JS::NullValue()"
+        if tag == self.jtc.BOOLEAN:
+            return "$JS::BooleanValue(%s)" % str(self.box.as_uint32() != 0).lower()
+        if tag == self.jtc.MAGIC:
+            value = self.box.as_uint32()
+            if 0 <= value and value < len(self.jtc.magic_names):
+                return "$JS::MagicValue(%s)" % (self.jtc.magic_names[value],)
+            else:
+                return "$JS::MagicValue(%d)" % (value,)
+
+        if tag == self.jtc.INT32:
+            value = self.box.as_uint32()
+            signbit = 1 << 31
+            value = (value ^ signbit) - signbit
+            return "$JS::Int32Value(%s)" % value
+
+        if tag == self.jtc.DOUBLE:
+            return "$JS::DoubleValue(%s)" % self.box.as_double()
+
+        if tag == self.jtc.STRING:
+            value = self.box.as_address().cast(self.cache.JSString_ptr_t)
+        elif tag == self.jtc.OBJECT:
+            value = self.box.as_address().cast(self.cache.JSObject_ptr_t)
+        elif tag == self.jtc.SYMBOL:
+            value = self.box.as_address().cast(self.cache.JSSymbol_ptr_t)
+        elif tag == self.jtc.BIGINT:
+            return "$JS::BigIntValue()"
+        else:
+            value = "unrecognized!"
+        return "$JS::Value(%s)" % (value,)
diff --git a/js/src/gdb/mozilla/prettyprinters.py b/js/src/gdb/mozilla/prettyprinters.py
new file mode 100644
index 0000000000..2296068936
--- /dev/null
+++ b/js/src/gdb/mozilla/prettyprinters.py
@@ -0,0 +1,446 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# mozilla/prettyprinters.py --- infrastructure for SpiderMonkey's auto-loaded pretty-printers.
+
+import gdb
+import re
+
+# Decorators for declaring pretty-printers.
+#
+# In each case, the decoratee should be a SpiderMonkey-style pretty-printer
+# factory, taking both a gdb.Value instance and a TypeCache instance as
+# arguments; see TypeCache, below.
+
+
+def check_for_reused_pretty_printer(fn):
+    # Check that |fn| hasn't been registered as a pretty-printer under some
+    # other name already. (The 'enabled' flags used by GDB's
+    # 'enable/disable/info pretty-printer' commands are simply stored as
+    # properties of the function objects themselves, so a single function
+    # object can't carry the 'enabled' flags for two different printers.)
+    if hasattr(fn, "enabled"):
+        raise RuntimeError("pretty-printer function %r registered more than once" % fn)
+
+
+# a dictionary mapping gdb.Type tags to pretty-printer functions.
+printers_by_tag = {}
+
+# A decorator: add the decoratee as a pretty-printer lookup function for types
+# named |type_name|.
+
+
+def pretty_printer(type_name):
+    def add(fn):
+        check_for_reused_pretty_printer(fn)
+        add_to_subprinter_list(fn, type_name)
+        printers_by_tag[type_name] = fn
+        return fn
+
+    return add
+
+
+# a dictionary mapping gdb.Type tags to pretty-printer functions for pointers to
+# that type.
+ptr_printers_by_tag = {}
+
+# A decorator: add the decoratee as a pretty-printer lookup function for
+# pointers to types named |type_name|.
+
+
+def ptr_pretty_printer(type_name):
+    def add(fn):
+        check_for_reused_pretty_printer(fn)
+        add_to_subprinter_list(fn, "ptr-to-" + type_name)
+        ptr_printers_by_tag[type_name] = fn
+        return fn
+
+    return add
+
+
+# a dictionary mapping gdb.Type tags to pretty-printer functions for
+# references to that type.
+ref_printers_by_tag = {}
+
+# A decorator: add the decoratee as a pretty-printer lookup function for
+# references to instances of types named |type_name|.
+
+
+def ref_pretty_printer(type_name):
+    def add(fn):
+        check_for_reused_pretty_printer(fn)
+        add_to_subprinter_list(fn, "ref-to-" + type_name)
+        ref_printers_by_tag[type_name] = fn
+        return fn
+
+    return add
+
+
+# a dictionary mapping the template name portion of gdb.Type tags to
+# pretty-printer functions for instantiations of that template.
+template_printers_by_tag = {}
+
+# A decorator: add the decoratee as a pretty-printer lookup function for
+# instantiations of templates named |template_name|.
+
+
+def template_pretty_printer(template_name):
+    def add(fn):
+        check_for_reused_pretty_printer(fn)
+        add_to_subprinter_list(fn, "instantiations-of-" + template_name)
+        template_printers_by_tag[template_name] = fn
+        return fn
+
+    return add
+
+
+# A list of (REGEXP, PRINTER) pairs, such that if REGEXP (a RegexObject)
+# matches the result of converting a gdb.Value's type to a string, then
+# PRINTER is a pretty-printer lookup function that will probably like that
+# value.
+printers_by_regexp = []
+
+# A decorator: add the decoratee as a pretty-printer factory for types
+# that, when converted to a string, match |pattern|. Use |name| as the
+# pretty-printer's name, when listing, enabling and disabling.
+
+
+def pretty_printer_for_regexp(pattern, name):
+    compiled = re.compile(pattern)
+
+    def add(fn):
+        check_for_reused_pretty_printer(fn)
+        add_to_subprinter_list(fn, name)
+        printers_by_regexp.append((compiled, fn))
+        return fn
+
+    return add
+
+
+# Forget all pretty-printer lookup functions defined in the module name
+# |module_name|, if any exist. Use this at the top of each pretty-printer
+# module like this:
+#
+#   clear_module_printers(__name__)
+
+
+def clear_module_printers(module_name):
+    global printers_by_tag, ptr_printers_by_tag, ref_printers_by_tag
+    global template_printers_by_tag, printers_by_regexp
+
+    # Remove all pretty-printers defined in the module named |module_name|
+    # from d.
+    def clear_dictionary(d):
+        # Walk the dictionary, building a list of keys whose entries we
+        # should remove. (It's not safe to delete entries from a dictionary
+        # while we're iterating over it.)
+        to_delete = []
+        for (k, v) in d.items():
+            if v.__module__ == module_name:
+                to_delete.append(k)
+                remove_from_subprinter_list(v)
+        for k in to_delete:
+            del d[k]
+
+    clear_dictionary(printers_by_tag)
+    clear_dictionary(ptr_printers_by_tag)
+    clear_dictionary(ref_printers_by_tag)
+    clear_dictionary(template_printers_by_tag)
+
+    # Iterate over printers_by_regexp, deleting entries from the given module.
+    new_list = []
+    for p in printers_by_regexp:
+        if p.__module__ == module_name:
+            remove_from_subprinter_list(p)
+        else:
+            new_list.append(p)
+    printers_by_regexp = new_list
+
+
+# Our subprinters array. The 'subprinters' attributes of all lookup
+# functions returned by lookup_for_objfile point to this array instance,
+# which we mutate as subprinters are added and removed.
+subprinters = []
+
+# Set up the 'name' and 'enabled' attributes on |subprinter|, and add it to our
+# list of all SpiderMonkey subprinters.
+
+
+def add_to_subprinter_list(subprinter, name):
+    subprinter.name = name
+    subprinter.enabled = True
+    subprinters.append(subprinter)
+
+
+# Remove |subprinter| from our list of all SpiderMonkey subprinters.
+
+
+def remove_from_subprinter_list(subprinter):
+    subprinters.remove(subprinter)
+
+
+# An exception class meaning, "This objfile has no SpiderMonkey in it."
+
+
+class NotSpiderMonkeyObjfileError(TypeError):
+    pass
+
+
+# TypeCache: a cache for frequently used information about an objfile.
+#
+# When a new SpiderMonkey objfile is loaded, we construct an instance of
+# this class for it. Then, whenever we construct a pretty-printer for some
+# gdb.Value, we also pass, as a second argument, the TypeCache for the
+# objfile to which that value's type belongs.
+#
+# if objfile doesn't seem to have SpiderMonkey code in it, the constructor
+# raises NotSpiderMonkeyObjfileError.
+#
+# Pretty-printer modules may add attributes to this to hold their own
+# cached values. Such attributes should be named mod_NAME, where the module
+# is named mozilla.NAME; for example, mozilla.JSString should store its
+# metadata in the TypeCache's mod_JSString attribute.
+
+
+class TypeCache(object):
+    def __init__(self, objfile):
+        self.objfile = objfile
+
+        # Unfortunately, the Python interface doesn't allow us to specify
+        # the objfile in whose scope lookups should occur. But simply
+        # knowing that we need to lookup the types afresh is probably
+        # enough.
+        self.void_t = gdb.lookup_type("void")
+        self.void_ptr_t = self.void_t.pointer()
+        self.uintptr_t = gdb.lookup_type("uintptr_t")
+        try:
+            self.JSString_ptr_t = gdb.lookup_type("JSString").pointer()
+            self.JSSymbol_ptr_t = gdb.lookup_type("JS::Symbol").pointer()
+            self.JSObject_ptr_t = gdb.lookup_type("JSObject").pointer()
+        except gdb.error:
+            raise NotSpiderMonkeyObjfileError
+
+        self.mod_GCCellPtr = None
+        self.mod_Interpreter = None
+        self.mod_JSObject = None
+        self.mod_JSOp = None
+        self.mod_JSString = None
+        self.mod_JS_Value = None
+        self.mod_ExecutableAllocator = None
+        self.mod_IonGraph = None
+
+
+# Yield a series of all the types that |t| implements, by following typedefs
+# and iterating over base classes. Specifically:
+# - |t| itself is the first value yielded.
+# - If we yield a typedef, we later yield its definition.
+# - If we yield a type with base classes, we later yield those base classes.
+# - If we yield a type with some base classes that are typedefs,
+#   we yield all the type's base classes before following the typedefs.
+#
+# This is a hokey attempt to order the implemented types by meaningfulness when
+# pretty-printed. Perhaps it is entirely misguided, and we should actually
+# collect all applicable pretty-printers, and then use some ordering on the
+# pretty-printers themselves.
+#
+# We may yield a type more than once (say, if it appears more than once in the
+# class hierarchy).
+
+
+def implemented_types(t):
+
+    # Yield all types that follow |t|.
+    def followers(t):
+        if t.code == gdb.TYPE_CODE_TYPEDEF:
+            yield t.target()
+            for t2 in followers(t.target()):
+                yield t2
+        elif is_struct_or_union(t):
+            base_classes = []
+            for f in t.fields():
+                if f.is_base_class:
+                    yield f.type
+                    base_classes.append(f.type)
+            for b in base_classes:
+                for t2 in followers(b):
+                    yield t2
+
+    yield t
+    for t2 in followers(t):
+        yield t2
+
+
+template_regexp = re.compile("([\w_:]+)<")
+
+
+def is_struct_or_union(t):
+    return t.code in (gdb.TYPE_CODE_STRUCT, gdb.TYPE_CODE_UNION)
+
+
+def is_struct_or_union_or_enum(t):
+    return t.code in (gdb.TYPE_CODE_STRUCT, gdb.TYPE_CODE_UNION, gdb.TYPE_CODE_ENUM)
+
+
+# Construct and return a pretty-printer lookup function for objfile, or
+# return None if the objfile doesn't contain SpiderMonkey code
+# (specifically, definitions for SpiderMonkey types).
+
+
+def lookup_for_objfile(objfile):
+    # Create a type cache for this objfile.
+    try:
+        cache = TypeCache(objfile)
+    except NotSpiderMonkeyObjfileError:
+        if gdb.parameter("verbose"):
+            gdb.write(
+                "objfile '%s' has no SpiderMonkey code; not registering pretty-printers\n"
+                % (objfile.filename,)
+            )
+        return None
+
+    # Return a pretty-printer for |value|, if we have one. This is the lookup
+    # function object we place in each gdb.Objfile's pretty-printers list, so it
+    # carries |name|, |enabled|, and |subprinters| attributes.
+    def lookup(value):
+        # If |table| has a pretty-printer for |tag|, apply it to |value|.
+        def check_table(table, tag):
+            if tag in table:
+                f = table[tag]
+                if f.enabled:
+                    return f(value, cache)
+            return None
+
+        def check_table_by_type_name(table, t):
+            if t.code == gdb.TYPE_CODE_TYPEDEF:
+                return check_table(table, str(t))
+            elif is_struct_or_union_or_enum(t) and t.tag:
+                return check_table(table, t.tag)
+            else:
+                return None
+
+        for t in implemented_types(value.type):
+            if t.code == gdb.TYPE_CODE_PTR:
+                for t2 in implemented_types(t.target()):
+                    p = check_table_by_type_name(ptr_printers_by_tag, t2)
+                    if p:
+                        return p
+            elif t.code == gdb.TYPE_CODE_REF:
+                for t2 in implemented_types(t.target()):
+                    p = check_table_by_type_name(ref_printers_by_tag, t2)
+                    if p:
+                        return p
+            else:
+                p = check_table_by_type_name(printers_by_tag, t)
+                if p:
+                    return p
+                if is_struct_or_union(t) and t.tag:
+                    m = template_regexp.match(t.tag)
+                    if m:
+                        p = check_table(template_printers_by_tag, m.group(1))
+                        if p:
+                            return p
+
+        # Failing that, look for a printer in printers_by_regexp. We have
+        # to scan the whole list, so regexp printers should be used
+        # sparingly.
+        s = str(value.type)
+        for (r, f) in printers_by_regexp:
+            if f.enabled:
+                m = r.match(s)
+                if m:
+                    p = f(value, cache)
+                    if p:
+                        return p
+
+        # No luck.
+        return None
+
+    # Give |lookup| the attributes expected of a pretty-printer with
+    # subprinters, for enabling and disabling.
+    lookup.name = "SpiderMonkey"
+    lookup.enabled = True
+    lookup.subprinters = subprinters
+
+    return lookup
+
+
+# A base class for pretty-printers for pointer values that handles null
+# pointers, by declining to construct a pretty-printer for them at all.
+# Derived classes may simply assume that self.value is non-null.
+#
+# To help share code, this class can also be used with reference types.
+#
+# This class provides the following methods, which subclasses are free to
+# override:
+#
+# __init__(self, value, cache): Save value and cache as properties by those names
+#     on the instance.
+#
+# to_string(self): format the type's name and address, as GDB would, and then
+#     call a 'summary' method (which the subclass must define) to produce a
+#     description of the referent.
+#
+#     Note that pretty-printers returning a 'string' display hint must not use
+#     this default 'to_string' method, as GDB will take everything it returns,
+#     including the type name and address, as string contents.
+
+
+class Pointer(object):
+    def __new__(cls, value, cache):
+        # Don't try to provide pretty-printers for NULL pointers.
+        if value.type.strip_typedefs().code == gdb.TYPE_CODE_PTR and value == 0:
+            return None
+        return super(Pointer, cls).__new__(cls)
+
+    def __init__(self, value, cache):
+        self.value = value
+        self.cache = cache
+
+    def to_string(self):
+        # See comment above.
+        assert not hasattr(self, "display_hint") or self.display_hint() != "string"
+        concrete_type = self.value.type.strip_typedefs()
+        if concrete_type.code == gdb.TYPE_CODE_PTR:
+            address = self.value.cast(self.cache.void_ptr_t)
+        elif concrete_type.code == gdb.TYPE_CODE_REF:
+            address = "@" + str(self.value.address.cast(self.cache.void_ptr_t))
+        else:
+            assert not "mozilla.prettyprinters.Pointer applied to bad value type"
+        try:
+            summary = self.summary()
+        except gdb.MemoryError as r:
+            summary = str(r)
+        v = "(%s) %s %s" % (self.value.type, address, summary)
+        return v
+
+    def summary(self):
+        raise NotImplementedError
+
+
+field_enum_value = None
+
+# Given |t|, a gdb.Type instance representing an enum type, return the
+# numeric value of the enum value named |name|.
+#
+# Pre-2012-4-18 versions of GDB store the value of an enum member on the
+# gdb.Field's 'bitpos' attribute; later versions store it on the 'enumval'
+# attribute. This function retrieves the value from either.
+
+
+def enum_value(t, name):
+    global field_enum_value
+    f = t[name]
+    # Monkey-patching is a-okay in polyfills! Just because.
+    if not field_enum_value:
+        if hasattr(f, "enumval"):
+
+            def field_enum_value(f):
+                return f.enumval
+
+        else:
+
+            def field_enum_value(f):
+                return f.bitpos
+
+    return field_enum_value(f)
diff --git a/js/src/gdb/mozilla/unwind.py b/js/src/gdb/mozilla/unwind.py
new file mode 100644
index 0000000000..8dd0ec8311
--- /dev/null
+++ b/js/src/gdb/mozilla/unwind.py
@@ -0,0 +1,596 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# mozilla/unwind.py --- unwinder and frame filter for SpiderMonkey
+
+import gdb
+import gdb.types
+from gdb.FrameDecorator import FrameDecorator
+import platform
+
+# For ease of use in Python 2, we use "long" instead of "int"
+# everywhere.
+try:
+    long
+except NameError:
+    long = int
+
+# The Python 3 |map| built-in works lazily, but in Python 2 we need
+# itertools.imap to get this.
+try:
+    from itertools import imap
+except ImportError:
+    imap = map
+
+_have_unwinder = True
+try:
+    from gdb.unwinder import Unwinder
+except ImportError:
+    _have_unwinder = False
+    # We need something here; it doesn't matter what as no unwinder
+    # will ever be instantiated.
+    Unwinder = object
+
+
+def debug(something):
+    # print("@@ " + something)
+    pass
+
+
+# Maps frametype enum base names to corresponding class.
+SizeOfFramePrefix = {
+    "FrameType::IonJS": "ExitFrameLayout",
+    "FrameType::BaselineJS": "JitFrameLayout",
+    "FrameType::BaselineStub": "BaselineStubFrameLayout",
+    "FrameType::IonStub": "JitStubFrameLayout",
+    "FrameType::CppToJSJit": "JitFrameLayout",
+    "FrameType::WasmToJSJit": "JitFrameLayout",
+    "FrameType::JSJitToWasm": "JitFrameLayout",
+    "FrameType::Rectifier": "RectifierFrameLayout",
+    "FrameType::IonAccessorIC": "IonAccessorICFrameLayout",
+    "FrameType::IonICCall": "IonICCallFrameLayout",
+    "FrameType::Exit": "ExitFrameLayout",
+    "FrameType::Bailout": "JitFrameLayout",
+}
+
+
+# We cannot have semi-colon as identifier names, so use a colon instead,
+# and forward the name resolution to the type cache class.
+class UnwinderTypeCacheFrameType(object):
+    def __init__(self, tc):
+        self.tc = tc
+
+    def __getattr__(self, name):
+        return self.tc.__getattr__("FrameType::" + name)
+
+
+class UnwinderTypeCache(object):
+    # All types and symbols that we need are attached to an object that we
+    # can dispose of as needed.
+
+    def __init__(self):
+        self.d = None
+        self.frame_enum_names = {}
+        self.frame_class_types = {}
+
+    # We take this bizarre approach to defer trying to look up any
+    # symbols until absolutely needed.  Without this, the loading
+    # approach taken by the gdb-tests would cause spurious exceptions.
+    def __getattr__(self, name):
+        if self.d is None:
+            self.initialize()
+        if name == "frame_type":
+            return UnwinderTypeCacheFrameType(self)
+        return self.d[name]
+
+    def value(self, name):
+        return long(gdb.lookup_symbol(name)[0].value())
+
+    def jit_value(self, name):
+        return self.value("js::jit::" + name)
+
+    def initialize(self):
+        self.d = {}
+        self.d["FRAMETYPE_MASK"] = (1 << self.jit_value("FRAMETYPE_BITS")) - 1
+        self.d["FRAMESIZE_SHIFT"] = self.jit_value("FRAMESIZE_SHIFT")
+        self.d["FRAME_HEADER_SIZE_SHIFT"] = self.jit_value("FRAME_HEADER_SIZE_SHIFT")
+        self.d["FRAME_HEADER_SIZE_MASK"] = self.jit_value("FRAME_HEADER_SIZE_MASK")
+
+        self.compute_frame_info()
+        commonFrameLayout = gdb.lookup_type("js::jit::CommonFrameLayout")
+        self.d["typeCommonFrameLayout"] = commonFrameLayout
+        self.d["typeCommonFrameLayoutPointer"] = commonFrameLayout.pointer()
+        self.d["per_tls_context"] = gdb.lookup_global_symbol("js::TlsContext")
+
+        self.d["void_starstar"] = gdb.lookup_type("void").pointer().pointer()
+
+        jitframe = gdb.lookup_type("js::jit::JitFrameLayout")
+        self.d["jitFrameLayoutPointer"] = jitframe.pointer()
+
+        self.d["CalleeToken_Function"] = self.jit_value("CalleeToken_Function")
+        self.d["CalleeToken_FunctionConstructing"] = self.jit_value(
+            "CalleeToken_FunctionConstructing"
+        )
+        self.d["CalleeToken_Script"] = self.jit_value("CalleeToken_Script")
+        self.d["JSFunction"] = gdb.lookup_type("JSFunction").pointer()
+        self.d["JSScript"] = gdb.lookup_type("JSScript").pointer()
+        self.d["Value"] = gdb.lookup_type("JS::Value")
+
+        self.d["SOURCE_SLOT"] = self.value("js::ScriptSourceObject::SOURCE_SLOT")
+        self.d["NativeObject"] = gdb.lookup_type("js::NativeObject").pointer()
+        self.d["HeapSlot"] = gdb.lookup_type("js::HeapSlot").pointer()
+        self.d["ScriptSource"] = gdb.lookup_type("js::ScriptSource").pointer()
+
+        # ProcessExecutableMemory, used to identify if a pc is in the section
+        # pre-allocated by the JIT.
+        self.d["MaxCodeBytesPerProcess"] = self.jit_value("MaxCodeBytesPerProcess")
+        self.d["execMemory"] = gdb.lookup_symbol("::execMemory")[0].value()
+
+    # Compute maps related to jit frames.
+    def compute_frame_info(self):
+        t = gdb.lookup_type("enum js::jit::FrameType")
+        for field in t.fields():
+            # Strip off "js::jit::", remains: "FrameType::*".
+            name = field.name[9:]
+            enumval = long(field.enumval)
+            self.d[name] = enumval
+            self.frame_enum_names[enumval] = name
+            class_type = gdb.lookup_type("js::jit::" + SizeOfFramePrefix[name])
+            self.frame_class_types[enumval] = class_type.pointer()
+
+
+class FrameSymbol(object):
+    "A symbol/value pair as expected from gdb frame decorators."
+
+    def __init__(self, sym, val):
+        self.sym = sym
+        self.val = val
+
+    def symbol(self):
+        return self.sym
+
+    def value(self):
+        return self.val
+
+
+class JitFrameDecorator(FrameDecorator):
+    """This represents a single JIT frame for the purposes of display.
+    That is, the frame filter creates instances of this when it sees a
+    JIT frame in the stack."""
+
+    def __init__(self, base, info, cache):
+        super(JitFrameDecorator, self).__init__(base)
+        self.info = info
+        self.cache = cache
+
+    def _decode_jitframe(self, this_frame):
+        calleetoken = long(this_frame["calleeToken_"])
+        tag = calleetoken & 3
+        calleetoken = calleetoken ^ tag
+        function = None
+        script = None
+        if (
+            tag == self.cache.CalleeToken_Function
+            or tag == self.cache.CalleeToken_FunctionConstructing
+        ):
+            fptr = gdb.Value(calleetoken).cast(self.cache.JSFunction)
+            try:
+                atom = fptr["atom_"]
+                if atom:
+                    function = str(atom)
+            except gdb.MemoryError:
+                function = "(could not read function name)"
+            script = fptr["u"]["scripted"]["s"]["script_"]
+        elif tag == self.cache.CalleeToken_Script:
+            script = gdb.Value(calleetoken).cast(self.cache.JSScript)
+        return {"function": function, "script": script}
+
+    def function(self):
+        if self.info["name"] is None:
+            return FrameDecorator.function(self)
+        name = self.info["name"]
+        result = "<<" + name
+        # If we have a frame, we can extract the callee information
+        # from it for display here.
+        this_frame = self.info["this_frame"]
+        if this_frame is not None:
+            if gdb.types.has_field(this_frame.type.target(), "calleeToken_"):
+                function = self._decode_jitframe(this_frame)["function"]
+                if function is not None:
+                    result = result + " " + function
+        return result + ">>"
+
+    def filename(self):
+        this_frame = self.info["this_frame"]
+        if this_frame is not None:
+            if gdb.types.has_field(this_frame.type.target(), "calleeToken_"):
+                script = self._decode_jitframe(this_frame)["script"]
+                if script is not None:
+                    obj = script["sourceObject_"]["value"]
+                    # Verify that this is a ScriptSource object.
+                    # FIXME should also deal with wrappers here.
+                    nativeobj = obj.cast(self.cache.NativeObject)
+                    # See bug 987069 and despair.  At least this
+                    # approach won't give exceptions.
+                    class_name = nativeobj["group_"]["value"]["clasp_"]["name"].string(
+                        "ISO-8859-1"
+                    )
+                    if class_name != "ScriptSource":
+                        return FrameDecorator.filename(self)
+                    scriptsourceobj = (nativeobj + 1).cast(self.cache.HeapSlot)[
+                        self.cache.SOURCE_SLOT
+                    ]
+                    scriptsource = scriptsourceobj["value"]["asBits_"] << 1
+                    scriptsource = scriptsource.cast(self.cache.ScriptSource)
+                    return scriptsource["filename_"]["mTuple"]["mFirstA"].string()
+        return FrameDecorator.filename(self)
+
+    def frame_args(self):
+        this_frame = self.info["this_frame"]
+        if this_frame is None:
+            return FrameDecorator.frame_args(self)
+        if not gdb.types.has_field(this_frame.type.target(), "numActualArgs_"):
+            return FrameDecorator.frame_args(self)
+        # See if this is a function call.
+        if self._decode_jitframe(this_frame)["function"] is None:
+            return FrameDecorator.frame_args(self)
+        # Construct and return an iterable of all the arguments.
+        result = []
+        num_args = long(this_frame["numActualArgs_"])
+        # Sometimes we see very large values here, so truncate it to
+        # bypass the damage.
+        if num_args > 10:
+            num_args = 10
+        args_ptr = (this_frame + 1).cast(self.cache.Value.pointer())
+        for i in range(num_args + 1):
+            # Synthesize names, since there doesn't seem to be
+            # anything better to do.
+            if i == 0:
+                name = "this"
+            else:
+                name = "arg%d" % i
+            result.append(FrameSymbol(name, args_ptr[i]))
+        return result
+
+
+class SpiderMonkeyFrameFilter(object):
+    "A frame filter for SpiderMonkey."
+
+    # |state_holder| is either None, or an instance of
+    # SpiderMonkeyUnwinder.  If the latter, then this class will
+    # reference the |unwinder_state| attribute to find the current
+    # unwinder state.
+    def __init__(self, cache, state_holder):
+        self.name = "SpiderMonkey"
+        self.enabled = True
+        self.priority = 100
+        self.state_holder = state_holder
+        self.cache = cache
+
+    def maybe_wrap_frame(self, frame):
+        if self.state_holder is None or self.state_holder.unwinder_state is None:
+            return frame
+        base = frame.inferior_frame()
+        info = self.state_holder.unwinder_state.get_frame(base)
+        if info is None:
+            return frame
+        return JitFrameDecorator(frame, info, self.cache)
+
+    def filter(self, frame_iter):
+        return imap(self.maybe_wrap_frame, frame_iter)
+
+
+class SpiderMonkeyFrameId(object):
+    "A frame id class, as specified by the gdb unwinder API."
+
+    def __init__(self, sp, pc):
+        self.sp = sp
+        self.pc = pc
+
+
+class UnwinderState(object):
+    """This holds all the state needed during a given unwind.  Each time a
+    new unwind is done, a new instance of this class is created.  It
+    keeps track of all the state needed to unwind JIT frames.  Note that
+    this class is not directly instantiated.
+
+    This is a base class, and must be specialized for each target
+    architecture, both because we need to use arch-specific register
+    names, and because entry frame unwinding is arch-specific.
+    See https://sourceware.org/bugzilla/show_bug.cgi?id=19286 for info
+    about the register name issue.
+
+    Each subclass must define SP_REGISTER, PC_REGISTER, and
+    SENTINEL_REGISTER (see x64UnwinderState for info); and implement
+    unwind_entry_frame_registers."""
+
+    def __init__(self, typecache):
+        self.next_sp = None
+        self.next_type = None
+        self.activation = None
+        # An unwinder instance is specific to a thread.  Record the
+        # selected thread for later verification.
+        self.thread = gdb.selected_thread()
+        self.frame_map = {}
+        self.typecache = typecache
+
+    # If the given gdb.Frame was created by this unwinder, return the
+    # corresponding informational dictionary for the frame.
+    # Otherwise, return None.  This is used by the frame filter to
+    # display extra information about the frame.
+    def get_frame(self, frame):
+        sp = long(frame.read_register(self.SP_REGISTER))
+        if sp in self.frame_map:
+            return self.frame_map[sp]
+        return None
+
+    # Add information about a frame to the frame map.  This map is
+    # queried by |self.get_frame|.  |sp| is the frame's stack pointer,
+    # and |name| the frame's type as a string, e.g. "FrameType::Exit".
+    def add_frame(self, sp, name=None, this_frame=None):
+        self.frame_map[long(sp)] = {"name": name, "this_frame": this_frame}
+
+    # See whether |pc| is claimed by the Jit.
+    def is_jit_address(self, pc):
+        execMem = self.typecache.execMemory
+        base = long(execMem["base_"])
+        length = self.typecache.MaxCodeBytesPerProcess
+
+        # If the base pointer is null, then no memory got allocated yet.
+        if long(base) == 0:
+            return False
+
+        # If allocated, then we allocated MaxCodeBytesPerProcess.
+        return base <= pc and pc < base + length
+
+    # Check whether |self| is valid for the selected thread.
+    def check(self):
+        return gdb.selected_thread() is self.thread
+
+    # Essentially js::TlsContext.get().
+    def get_tls_context(self):
+        return self.typecache.per_tls_context.value()["mValue"]
+
+    # |common| is a pointer to a CommonFrameLayout object.  Return a
+    # tuple (local_size, header_size, frame_type), where |size| is the
+    # integer size of the previous frame's locals; |header_size| is
+    # the size of this frame's header; and |frame_type| is an integer
+    # representing the previous frame's type.
+    def unpack_descriptor(self, common):
+        value = long(common["descriptor_"])
+        local_size = value >> self.typecache.FRAMESIZE_SHIFT
+        header_size = (
+            value >> self.typecache.FRAME_HEADER_SIZE_SHIFT
+        ) & self.typecache.FRAME_HEADER_SIZE_MASK
+        header_size = header_size * self.typecache.void_starstar.sizeof
+        frame_type = long(value & self.typecache.FRAMETYPE_MASK)
+        if frame_type == self.typecache.frame_type.CppToJSJit:
+            # Trampoline-x64.cpp pushes a JitFrameLayout object, but
+            # the stack pointer is actually adjusted as if a
+            # CommonFrameLayout object was pushed.
+            header_size = self.typecache.typeCommonFrameLayout.sizeof
+        return (local_size, header_size, frame_type)
+
+    # Create a new frame for gdb.  This makes a new unwind info object
+    # and fills it in, then returns it.  It also registers any
+    # pertinent information with the frame filter for later display.
+    #
+    # |pc| is the PC from the pending frame
+    # |sp| is the stack pointer to use
+    # |frame| points to the CommonFrameLayout object
+    # |frame_type| is a integer, one of the |enum FrameType| values,
+    # describing the current frame.
+    # |pending_frame| is the pending frame (see the gdb unwinder
+    # documentation).
+    def create_frame(self, pc, sp, frame, frame_type, pending_frame):
+        # Make a frame_id that claims that |frame| is sort of like a
+        # frame pointer for this frame.
+        frame_id = SpiderMonkeyFrameId(frame, pc)
+
+        # Read the frame layout object to find the next such object.
+        # This lets us unwind the necessary registers for the next
+        # frame, and also update our internal state to match.
+        common = frame.cast(self.typecache.typeCommonFrameLayoutPointer)
+        next_pc = common["returnAddress_"]
+        (local_size, header_size, next_type) = self.unpack_descriptor(common)
+        next_sp = frame + header_size + local_size
+
+        # Compute the type of the next oldest frame's descriptor.
+        this_class_type = self.typecache.frame_class_types[frame_type]
+        this_frame = frame.cast(this_class_type)
+
+        # Register this frame so the frame filter can find it.  This
+        # is registered using SP because we don't have any other good
+        # approach -- you can't get the frame id from a gdb.Frame.
+        # https://sourceware.org/bugzilla/show_bug.cgi?id=19800
+        frame_name = self.typecache.frame_enum_names[frame_type]
+        self.add_frame(sp, name=frame_name, this_frame=this_frame)
+
+        # Update internal state for the next unwind.
+        self.next_sp = next_sp
+        self.next_type = next_type
+
+        unwind_info = pending_frame.create_unwind_info(frame_id)
+        unwind_info.add_saved_register(self.PC_REGISTER, next_pc)
+        unwind_info.add_saved_register(self.SP_REGISTER, next_sp)
+        # FIXME it would be great to unwind any other registers here.
+        return unwind_info
+
+    # Unwind an "ordinary" JIT frame.  This is used for JIT frames
+    # other than enter and exit frames.  Returns the newly-created
+    # unwind info for gdb.
+    def unwind_ordinary(self, pc, pending_frame):
+        return self.create_frame(
+            pc, self.next_sp, self.next_sp, self.next_type, pending_frame
+        )
+
+    # Unwind an exit frame.  Returns None if this cannot be done;
+    # otherwise returns the newly-created unwind info for gdb.
+    def unwind_exit_frame(self, pc, pending_frame):
+        if self.activation == 0:
+            # Reached the end of the list.
+            return None
+        elif self.activation is None:
+            cx = self.get_tls_context()
+            self.activation = cx["jitActivation"]["value"]
+        else:
+            self.activation = self.activation["prevJitActivation_"]
+
+        packedExitFP = self.activation["packedExitFP_"]
+        if packedExitFP == 0:
+            return None
+
+        exit_sp = pending_frame.read_register(self.SP_REGISTER)
+        frame_type = self.typecache.frame_type.Exit
+        return self.create_frame(pc, exit_sp, packedExitFP, frame_type, pending_frame)
+
+    # A wrapper for unwind_entry_frame_registers that handles
+    # architecture-independent boilerplate.
+    def unwind_entry_frame(self, pc, pending_frame):
+        sp = self.next_sp
+        # Notify the frame filter.
+        self.add_frame(sp, name="FrameType::CppToJSJit")
+        # Make an unwind_info for the per-architecture code to fill in.
+        frame_id = SpiderMonkeyFrameId(sp, pc)
+        unwind_info = pending_frame.create_unwind_info(frame_id)
+        self.unwind_entry_frame_registers(sp, unwind_info)
+        self.next_sp = None
+        self.next_type = None
+        return unwind_info
+
+    # The main entry point that is called to try to unwind a JIT frame
+    # of any type.  Returns None if this cannot be done; otherwise
+    # returns the newly-created unwind info for gdb.
+    def unwind(self, pending_frame):
+        pc = pending_frame.read_register(self.PC_REGISTER)
+
+        # If the jit does not claim this address, bail.  GDB defers to our
+        # unwinder by default, but we don't really want that kind of power.
+        if not self.is_jit_address(long(pc)):
+            return None
+
+        if self.next_sp is not None:
+            if self.next_type == self.typecache.frame_type.CppToJSJit:
+                return self.unwind_entry_frame(pc, pending_frame)
+            return self.unwind_ordinary(pc, pending_frame)
+        # Maybe we've found an exit frame.  FIXME I currently don't
+        # know how to identify these precisely, so we'll just hope for
+        # the time being.
+        return self.unwind_exit_frame(pc, pending_frame)
+
+
+class x64UnwinderState(UnwinderState):
+    "The UnwinderState subclass for x86-64."
+
+    SP_REGISTER = "rsp"
+    PC_REGISTER = "rip"
+
+    # A register unique to this architecture, that is also likely to
+    # have been saved in any frame.  The best thing to use here is
+    # some arch-specific name for PC or SP.
+    SENTINEL_REGISTER = "rip"
+
+    # Must be in sync with Trampoline-x64.cpp:generateEnterJIT.  Note
+    # that rip isn't pushed there explicitly, but rather by the
+    # previous function's call.
+    PUSHED_REGS = ["r15", "r14", "r13", "r12", "rbx", "rbp", "rip"]
+
+    # Fill in the unwound registers for an entry frame.
+    def unwind_entry_frame_registers(self, sp, unwind_info):
+        sp = sp.cast(self.typecache.void_starstar)
+        # Skip the "result" push.
+        sp = sp + 1
+        for reg in self.PUSHED_REGS:
+            data = sp.dereference()
+            sp = sp + 1
+            unwind_info.add_saved_register(reg, data)
+            if reg == "rbp":
+                unwind_info.add_saved_register(self.SP_REGISTER, sp)
+
+
+class SpiderMonkeyUnwinder(Unwinder):
+    """The unwinder object.  This provides the "user interface" to the JIT
+    unwinder, and also handles constructing or destroying UnwinderState
+    objects as needed."""
+
+    # A list of all the possible unwinders.  See |self.make_unwinder|.
+    UNWINDERS = [x64UnwinderState]
+
+    def __init__(self, typecache):
+        super(SpiderMonkeyUnwinder, self).__init__("SpiderMonkey")
+        self.typecache = typecache
+        self.unwinder_state = None
+
+        # Disabled by default until we figure out issues in gdb.
+        self.enabled = False
+        gdb.write(
+            "SpiderMonkey unwinder is disabled by default, to enable it type:\n"
+            + "\tenable unwinder .* SpiderMonkey\n"
+        )
+        # Some versions of gdb did not flush the internal frame cache
+        # when enabling or disabling an unwinder.  This was fixed in
+        # the same release of gdb that added the breakpoint_created
+        # event.
+        if not hasattr(gdb.events, "breakpoint_created"):
+            gdb.write("\tflushregs\n")
+
+        # We need to invalidate the unwinder state whenever the
+        # inferior starts executing.  This avoids having a stale
+        # cache.
+        gdb.events.cont.connect(self.invalidate_unwinder_state)
+        assert self.test_sentinels()
+
+    def test_sentinels(self):
+        # Self-check.
+        regs = {}
+        for unwinder in self.UNWINDERS:
+            if unwinder.SENTINEL_REGISTER in regs:
+                return False
+            regs[unwinder.SENTINEL_REGISTER] = 1
+        return True
+
+    def make_unwinder(self, pending_frame):
+        # gdb doesn't provide a good way to find the architecture.
+        # See https://sourceware.org/bugzilla/show_bug.cgi?id=19399
+        # So, we look at each known architecture and see if the
+        # corresponding "unique register" is known.
+        for unwinder in self.UNWINDERS:
+            try:
+                pending_frame.read_register(unwinder.SENTINEL_REGISTER)
+            except Exception:
+                # Failed to read the register, so let's keep going.
+                # This is more fragile than it might seem, because it
+                # fails if the sentinel register wasn't saved in the
+                # previous frame.
+                continue
+            return unwinder(self.typecache)
+        return None
+
+    def __call__(self, pending_frame):
+        if self.unwinder_state is None or not self.unwinder_state.check():
+            self.unwinder_state = self.make_unwinder(pending_frame)
+        if not self.unwinder_state:
+            return None
+        return self.unwinder_state.unwind(pending_frame)
+
+    def invalidate_unwinder_state(self, *args, **kwargs):
+        self.unwinder_state = None
+
+
+def register_unwinder(objfile):
+    """Register the unwinder and frame filter with |objfile|.  If |objfile|
+    is None, register them globally."""
+
+    type_cache = UnwinderTypeCache()
+    unwinder = None
+    # This currently only works on Linux, due to parse_proc_maps.
+    if _have_unwinder and platform.system() == "Linux":
+        unwinder = SpiderMonkeyUnwinder(type_cache)
+        gdb.unwinder.register_unwinder(objfile, unwinder, replace=True)
+    # We unconditionally register the frame filter, because at some
+    # point we'll add interpreter frame filtering.
+    filt = SpiderMonkeyFrameFilter(type_cache, unwinder)
+    if objfile is None:
+        objfile = gdb
+    objfile.frame_filters[filt.name] = filt
diff --git a/js/src/gdb/progressbar.py b/js/src/gdb/progressbar.py
new file mode 100644
index 0000000000..39c3628b18
--- /dev/null
+++ b/js/src/gdb/progressbar.py
@@ -0,0 +1,54 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this
+# file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# Text progress bar library, like curl or scp.
+
+import sys
+import datetime
+import time
+
+
+class ProgressBar(object):
+    def __init__(self, label, limit, label_width=12):
+        self.label = label
+        self.limit = limit
+        self.label_width = label_width
+        self.cur = 0
+        self.t0 = datetime.datetime.now()
+        self.fullwidth = None
+
+        self.barlen = 64 - self.label_width
+        self.fmt = (
+            "\r%-" + str(label_width) + "s %3d%% %-" + str(self.barlen) + "s| %6.1fs"
+        )
+
+    def update(self, value):
+        self.cur = value
+        pct = int(100.0 * self.cur / self.limit)
+        barlen = int(1.0 * self.barlen * self.cur / self.limit) - 1
+        bar = "=" * barlen + ">"
+        dt = datetime.datetime.now() - self.t0
+        dt = dt.seconds + dt.microseconds * 1e-6
+        line = self.fmt % (self.label[: self.label_width], pct, bar, dt)
+        self.fullwidth = len(line)
+        sys.stdout.write(line)
+        sys.stdout.flush()
+
+    # Clear the current bar and leave the cursor at the start of the line.
+    def clear(self):
+        if self.fullwidth:
+            sys.stdout.write("\r" + " " * self.fullwidth + "\r")
+            self.fullwidth = None
+
+    def finish(self):
+        self.update(self.limit)
+        sys.stdout.write("\n")
+
+
+if __name__ == "__main__":
+    pb = ProgressBar("test", 12)
+    for i in range(12):
+        pb.update(i)
+        time.sleep(0.5)
+    pb.finish()
diff --git a/js/src/gdb/run-tests.py b/js/src/gdb/run-tests.py
new file mode 100644
index 0000000000..da8d38cfc1
--- /dev/null
+++ b/js/src/gdb/run-tests.py
@@ -0,0 +1,470 @@
+#!/usr/bin/env python
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this
+# file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# run-tests.py -- Python harness for GDB SpiderMonkey support
+
+import os
+import re
+import subprocess
+import sys
+
+# From this directory:
+import progressbar
+from taskpool import TaskPool, get_cpu_count
+
+
+def _relpath(path, start=None):
+    # Backported from Python 3.1 posixpath.py
+
+    """Return a relative version of a path"""
+
+    if not path:
+        raise ValueError("no path specified")
+
+    if start is None:
+        start = os.curdir
+
+    start_list = os.path.abspath(start).split(os.sep)
+    path_list = os.path.abspath(path).split(os.sep)
+
+    # Work out how much of the filepath is shared by start and path.
+    i = len(os.path.commonprefix([start_list, path_list]))
+
+    rel_list = [os.pardir] * (len(start_list) - i) + path_list[i:]
+    if not rel_list:
+        return os.curdir
+    return os.path.join(*rel_list)
+
+
+os.path.relpath = _relpath
+
+# Characters that need to be escaped when used in shell words.
+shell_need_escapes = re.compile("[^\w\d%+,-./:=@'\"]", re.DOTALL)
+# Characters that need to be escaped within double-quoted strings.
+shell_dquote_escapes = re.compile('[^\w\d%+,-./:=@"]', re.DOTALL)
+
+
+def make_shell_cmd(l):
+    def quote(s):
+        if shell_need_escapes.search(s):
+            if s.find("'") < 0:
+                return "'" + s + "'"
+            return '"' + shell_dquote_escapes.sub("\\g<0>", s) + '"'
+        return s
+
+    return " ".join([quote(_) for _ in l])
+
+
+# An instance of this class collects the lists of passing, failing, and
+# timing-out tests, runs the progress bar, and prints a summary at the end.
+class Summary(object):
+    class SummaryBar(progressbar.ProgressBar):
+        def __init__(self, limit):
+            super(Summary.SummaryBar, self).__init__("", limit, 24)
+
+        def start(self):
+            self.label = "[starting           ]"
+            self.update(0)
+
+        def counts(self, run, failures, timeouts):
+            self.label = "[%4d|%4d|%4d|%4d]" % (run - failures, failures, timeouts, run)
+            self.update(run)
+
+    def __init__(self, num_tests):
+        self.run = 0
+        self.failures = []  # kind of judgemental; "unexpecteds"?
+        self.timeouts = []
+        if not OPTIONS.hide_progress:
+            self.bar = Summary.SummaryBar(num_tests)
+
+    # Progress bar control.
+    def start(self):
+        if not OPTIONS.hide_progress:
+            self.bar.start()
+
+    def update(self):
+        if not OPTIONS.hide_progress:
+            self.bar.counts(self.run, len(self.failures), len(self.timeouts))
+
+    # Call 'thunk' to show some output, while getting the progress bar out of the way.
+
+    def interleave_output(self, thunk):
+        if not OPTIONS.hide_progress:
+            self.bar.clear()
+        thunk()
+        self.update()
+
+    def passed(self, test):
+        self.run += 1
+        self.update()
+
+    def failed(self, test):
+        self.run += 1
+        self.failures.append(test)
+        self.update()
+
+    def timeout(self, test):
+        self.run += 1
+        self.timeouts.append(test)
+        self.update()
+
+    def finish(self):
+        if not OPTIONS.hide_progress:
+            self.bar.finish()
+
+        if self.failures:
+
+            print("tests failed:")
+            for test in self.failures:
+                test.show(sys.stdout)
+
+            if OPTIONS.worklist:
+                try:
+                    with open(OPTIONS.worklist) as out:
+                        for test in self.failures:
+                            out.write(test.name + "\n")
+                except IOError as err:
+                    sys.stderr.write(
+                        "Error writing worklist file '%s': %s" % (OPTIONS.worklist, err)
+                    )
+                    sys.exit(1)
+
+            if OPTIONS.write_failures:
+                try:
+                    with open(OPTIONS.write_failures, "w") as out:
+                        for test in self.failures:
+                            test.show(out)
+                except IOError as err:
+                    sys.stderr.write(
+                        "Error writing worklist file '%s': %s"
+                        % (OPTIONS.write_failures, err)
+                    )
+                    sys.exit(1)
+
+        if self.timeouts:
+            print("tests timed out:")
+            for test in self.timeouts:
+                test.show(sys.stdout)
+
+        if self.failures or self.timeouts:
+            sys.exit(2)
+
+
+class Test(TaskPool.Task):
+    def __init__(self, path, summary):
+        super(Test, self).__init__()
+        self.test_path = path  # path to .py test file
+        self.summary = summary
+
+        # test.name is the name of the test relative to the top of the test
+        # directory. This is what we use to report failures and timeouts,
+        # and when writing test lists.
+        self.name = os.path.relpath(self.test_path, OPTIONS.testdir)
+
+        self.stdout = ""
+        self.stderr = ""
+        self.returncode = None
+
+    def cmd(self):
+        testlibdir = os.path.normpath(
+            os.path.join(OPTIONS.testdir, "..", "lib-for-tests")
+        )
+        return [
+            OPTIONS.gdb_executable,
+            "-nw",  # Don't create a window (unnecessary?)
+            "-nx",  # Don't read .gdbinit.
+            "--ex",
+            "add-auto-load-safe-path %s" % (OPTIONS.bindir,),
+            "--ex",
+            "set env LD_LIBRARY_PATH %s" % (OPTIONS.bindir,),
+            "--ex",
+            "file %s" % (os.path.join(OPTIONS.bindir, "gdb-tests"),),
+            "--eval-command",
+            "python testlibdir=%r" % (testlibdir,),
+            "--eval-command",
+            "python testscript=%r" % (self.test_path,),
+            "--eval-command",
+            "python exec(open(%r).read())" % os.path.join(testlibdir, "catcher.py"),
+        ]
+
+    def start(self, pipe, deadline):
+        super(Test, self).start(pipe, deadline)
+        if OPTIONS.show_cmd:
+            self.summary.interleave_output(lambda: self.show_cmd(sys.stdout))
+
+    def onStdout(self, text):
+        self.stdout += text
+
+    def onStderr(self, text):
+        self.stderr += text
+
+    def onFinished(self, returncode):
+        self.returncode = returncode
+        if OPTIONS.show_output:
+            self.summary.interleave_output(lambda: self.show_output(sys.stdout))
+        if returncode != 0:
+            self.summary.failed(self)
+        else:
+            self.summary.passed(self)
+
+    def onTimeout(self):
+        self.summary.timeout(self)
+
+    def show_cmd(self, out):
+        out.write("Command: %s\n" % (make_shell_cmd(self.cmd()),))
+
+    def show_output(self, out):
+        if self.stdout:
+            out.write("Standard output:")
+            out.write("\n" + self.stdout + "\n")
+        if self.stderr:
+            out.write("Standard error:")
+            out.write("\n" + self.stderr + "\n")
+
+    def show(self, out):
+        out.write(self.name + "\n")
+        if OPTIONS.write_failure_output:
+            self.show_cmd(out)
+            self.show_output(out)
+            out.write("GDB exit code: %r\n" % (self.returncode,))
+
+
+def find_tests(dir, substring=None):
+    ans = []
+    for dirpath, _, filenames in os.walk(dir):
+        if dirpath == ".":
+            continue
+        for filename in filenames:
+            if not filename.endswith(".py"):
+                continue
+            test = os.path.join(dirpath, filename)
+            if substring is None or substring in os.path.relpath(test, dir):
+                ans.append(test)
+    return ans
+
+
+def build_test_exec(builddir):
+    subprocess.check_call(["make"], cwd=builddir)
+
+
+def run_tests(tests, summary):
+    jobs = OPTIONS.workercount
+    # python 3.3 fixed a bug with concurrently writing .pyc files.
+    # https://bugs.python.org/issue13146
+    embedded_version = (
+        subprocess.check_output(
+            [
+                OPTIONS.gdb_executable,
+                "--batch",
+                "--ex",
+                "python import sys; print(sys.hexversion)",
+            ]
+        )
+        .decode("ascii")
+        .strip()
+    )
+    if hex(int(embedded_version)) < "0x3030000":
+        jobs = 1
+
+    pool = TaskPool(tests, job_limit=jobs, timeout=OPTIONS.timeout)
+    pool.run_all()
+
+
+OPTIONS = None
+
+
+def main(argv):
+    global OPTIONS
+    script_path = os.path.abspath(__file__)
+    script_dir = os.path.dirname(script_path)
+
+    # OBJDIR is a standalone SpiderMonkey build directory. This is where we
+    # find the SpiderMonkey shared library to link against.
+    #
+    # The [TESTS] optional arguments are paths of test files relative
+    # to the jit-test/tests directory.
+    from optparse import OptionParser
+
+    op = OptionParser(usage="%prog [options] OBJDIR [TESTS...]")
+    op.add_option(
+        "-s",
+        "--show-cmd",
+        dest="show_cmd",
+        action="store_true",
+        help="show GDB shell command run",
+    )
+    op.add_option(
+        "-o",
+        "--show-output",
+        dest="show_output",
+        action="store_true",
+        help="show output from GDB",
+    )
+    op.add_option(
+        "-x",
+        "--exclude",
+        dest="exclude",
+        action="append",
+        help="exclude given test dir or path",
+    )
+    op.add_option(
+        "-t",
+        "--timeout",
+        dest="timeout",
+        type=float,
+        default=150.0,
+        help="set test timeout in seconds",
+    )
+    op.add_option(
+        "-j",
+        "--worker-count",
+        dest="workercount",
+        type=int,
+        help="Run [WORKERCOUNT] tests at a time",
+    )
+    op.add_option(
+        "--no-progress",
+        dest="hide_progress",
+        action="store_true",
+        help="hide progress bar",
+    )
+    op.add_option(
+        "--worklist",
+        dest="worklist",
+        metavar="FILE",
+        help="Read tests to run from [FILE] (or run all if [FILE] not found);\n"
+        "write failures back to [FILE]",
+    )
+    op.add_option(
+        "-r",
+        "--read-tests",
+        dest="read_tests",
+        metavar="FILE",
+        help="Run test files listed in [FILE]",
+    )
+    op.add_option(
+        "-w",
+        "--write-failures",
+        dest="write_failures",
+        metavar="FILE",
+        help="Write failing tests to [FILE]",
+    )
+    op.add_option(
+        "--write-failure-output",
+        dest="write_failure_output",
+        action="store_true",
+        help="With --write-failures=FILE, additionally write the output of failed "
+        "tests to [FILE]",
+    )
+    op.add_option(
+        "--gdb",
+        dest="gdb_executable",
+        metavar="EXECUTABLE",
+        default="gdb",
+        help="Run tests with [EXECUTABLE], rather than plain 'gdb'.",
+    )
+    op.add_option(
+        "--srcdir",
+        dest="srcdir",
+        default=os.path.abspath(os.path.join(script_dir, "..")),
+        help="Use SpiderMonkey sources in [SRCDIR].",
+    )
+    op.add_option(
+        "--testdir",
+        dest="testdir",
+        default=os.path.join(script_dir, "tests"),
+        help="Find tests in [TESTDIR].",
+    )
+    op.add_option(
+        "--builddir", dest="builddir", help="Build test executable from [BUILDDIR]."
+    )
+    op.add_option("--bindir", dest="bindir", help="Run test executable from [BINDIR].")
+    (OPTIONS, args) = op.parse_args(argv)
+    if len(args) < 1:
+        op.error("missing OBJDIR argument")
+    OPTIONS.objdir = os.path.abspath(args[0])
+
+    test_args = args[1:]
+
+    if not OPTIONS.workercount:
+        OPTIONS.workercount = get_cpu_count()
+
+    # Compute defaults for OPTIONS.builddir and OPTIONS.bindir now, since we've
+    # computed OPTIONS.objdir.
+    if not OPTIONS.builddir:
+        OPTIONS.builddir = os.path.join(OPTIONS.objdir, "js", "src", "gdb")
+    if not OPTIONS.bindir:
+        OPTIONS.bindir = os.path.join(OPTIONS.objdir, "dist", "bin")
+
+    test_set = set()
+
+    # All the various sources of test names accumulate.
+    if test_args:
+        for arg in test_args:
+            test_set.update(find_tests(OPTIONS.testdir, arg))
+    if OPTIONS.worklist:
+        try:
+            with open(OPTIONS.worklist) as f:
+                for line in f:
+                    test_set.update(os.path.join(OPTIONS.testdir, line.strip("\n")))
+        except IOError:
+            # With worklist, a missing file means to start the process with
+            # the complete list of tests.
+            sys.stderr.write(
+                "Couldn't read worklist file '%s'; running all tests\n"
+                % (OPTIONS.worklist,)
+            )
+            test_set = set(find_tests(OPTIONS.testdir))
+    if OPTIONS.read_tests:
+        try:
+            with open(OPTIONS.read_tests) as f:
+                for line in f:
+                    test_set.update(os.path.join(OPTIONS.testdir, line.strip("\n")))
+        except IOError as err:
+            sys.stderr.write(
+                "Error trying to read test file '%s': %s\n" % (OPTIONS.read_tests, err)
+            )
+            sys.exit(1)
+
+    # If none of the above options were passed, and no tests were listed
+    # explicitly, use the complete set.
+    if not test_args and not OPTIONS.worklist and not OPTIONS.read_tests:
+        test_set = set(find_tests(OPTIONS.testdir))
+
+    if OPTIONS.exclude:
+        exclude_set = set()
+        for exclude in OPTIONS.exclude:
+            exclude_set.update(find_tests(OPTIONS.testdir, exclude))
+        test_set -= exclude_set
+
+    if not test_set:
+        sys.stderr.write("No tests found matching command line arguments.\n")
+        sys.exit(1)
+
+    summary = Summary(len(test_set))
+    test_list = [Test(_, summary) for _ in sorted(test_set)]
+
+    # Build the test executable from all the .cpp files found in the test
+    # directory tree.
+    try:
+        build_test_exec(OPTIONS.builddir)
+    except subprocess.CalledProcessError as err:
+        sys.stderr.write("Error building test executable: %s\n" % (err,))
+        sys.exit(1)
+
+    # Run the tests.
+    try:
+        summary.start()
+        run_tests(test_list, summary)
+        summary.finish()
+    except OSError as err:
+        sys.stderr.write("Error running tests: %s\n" % (err,))
+        sys.exit(1)
+
+    sys.exit(0)
+
+
+if __name__ == "__main__":
+    main(sys.argv[1:])
diff --git a/js/src/gdb/taskpool.py b/js/src/gdb/taskpool.py
new file mode 100644
index 0000000000..363031183d
--- /dev/null
+++ b/js/src/gdb/taskpool.py
@@ -0,0 +1,240 @@
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this file,
+# You can obtain one at http://mozilla.org/MPL/2.0/.
+
+# flake8: noqa: F821
+
+import fcntl
+import os
+import select
+import time
+from subprocess import Popen, PIPE
+
+
+class TaskPool(object):
+    # Run a series of subprocesses. Try to keep up to a certain number going in
+    # parallel at any given time. Enforce time limits.
+    #
+    # This is implemented using non-blocking I/O, and so is Unix-specific.
+    #
+    # We assume that, if a task closes its standard error, then it's safe to
+    # wait for it to terminate. So an ill-behaved task that closes its standard
+    # output and then hangs will hang us, as well. However, as it takes special
+    # effort to close one's standard output, this seems unlikely to be a
+    # problem in practice.
+
+    # A task we should run in a subprocess. Users should subclass this and
+    # fill in the methods as given.
+    class Task(object):
+        def __init__(self):
+            self.pipe = None
+            self.start_time = None
+
+        # Record that this task is running, with |pipe| as its Popen object,
+        # and should time out at |deadline|.
+        def start(self, pipe, deadline):
+            self.pipe = pipe
+            self.deadline = deadline
+
+        # Return a shell command (a string or sequence of arguments) to be
+        # passed to Popen to run the task. The command will be given
+        # /dev/null as its standard input, and pipes as its standard output
+        # and error.
+        def cmd(self):
+            raise NotImplementedError
+
+        # TaskPool calls this method to report that the process wrote
+        # |string| to its standard output.
+        def onStdout(self, string):
+            raise NotImplementedError
+
+        # TaskPool calls this method to report that the process wrote
+        # |string| to its standard error.
+        def onStderr(self, string):
+            raise NotImplementedError
+
+        # TaskPool calls this method to report that the process terminated,
+        # yielding |returncode|.
+        def onFinished(self, returncode):
+            raise NotImplementedError
+
+        # TaskPool calls this method to report that the process timed out and
+        # was killed.
+        def onTimeout(self):
+            raise NotImplementedError
+
+    # If a task output handler (onStdout, onStderr) throws this, we terminate
+    # the task.
+    class TerminateTask(Exception):
+        pass
+
+    def __init__(self, tasks, cwd=".", job_limit=4, timeout=150):
+        self.pending = iter(tasks)
+        self.cwd = cwd
+        self.job_limit = job_limit
+        self.timeout = timeout
+        self.next_pending = next(self.pending, None)
+
+    def run_all(self):
+        # The currently running tasks: a set of Task instances.
+        running = set()
+        with open(os.devnull, "r") as devnull:
+            while True:
+                while len(running) < self.job_limit and self.next_pending:
+                    task = self.next_pending
+                    p = Popen(
+                        task.cmd(),
+                        bufsize=16384,
+                        stdin=devnull,
+                        stdout=PIPE,
+                        stderr=PIPE,
+                        cwd=self.cwd,
+                    )
+
+                    # Put the stdout and stderr pipes in non-blocking mode. See
+                    # the post-'select' code below for details.
+                    flags = fcntl.fcntl(p.stdout, fcntl.F_GETFL)
+                    fcntl.fcntl(p.stdout, fcntl.F_SETFL, flags | os.O_NONBLOCK)
+                    flags = fcntl.fcntl(p.stderr, fcntl.F_GETFL)
+                    fcntl.fcntl(p.stderr, fcntl.F_SETFL, flags | os.O_NONBLOCK)
+
+                    task.start(p, time.time() + self.timeout)
+                    running.add(task)
+                    self.next_pending = next(self.pending, None)
+
+                # If we have no tasks running, and the above wasn't able to
+                # start any new ones, then we must be done!
+                if not running:
+                    break
+
+                # How many seconds do we have until the earliest deadline?
+                now = time.time()
+                secs_to_next_deadline = max(min([t.deadline for t in running]) - now, 0)
+
+                # Wait for output or a timeout.
+                stdouts_and_stderrs = [t.pipe.stdout for t in running] + [
+                    t.pipe.stderr for t in running
+                ]
+                (readable, w, x) = select.select(
+                    stdouts_and_stderrs, [], [], secs_to_next_deadline
+                )
+                finished = set()
+                terminate = set()
+                for t in running:
+                    # Since we've placed the pipes in non-blocking mode, these
+                    # 'read's will simply return as many bytes as are available,
+                    # rather than blocking until they have accumulated the full
+                    # amount requested (or reached EOF). The 'read's should
+                    # never throw, since 'select' has told us there was
+                    # something available.
+                    if t.pipe.stdout in readable:
+                        output = t.pipe.stdout.read(16384)
+                        if len(output):
+                            try:
+                                t.onStdout(output.decode("utf-8"))
+                            except TerminateTask:
+                                terminate.add(t)
+                    if t.pipe.stderr in readable:
+                        output = t.pipe.stderr.read(16384)
+                        if len(output):
+                            try:
+                                t.onStderr(output.decode("utf-8"))
+                            except TerminateTask:
+                                terminate.add(t)
+                        else:
+                            # We assume that, once a task has closed its stderr,
+                            # it will soon terminate. If a task closes its
+                            # stderr and then hangs, we'll hang too, here.
+                            t.pipe.wait()
+                            t.onFinished(t.pipe.returncode)
+                            finished.add(t)
+                # Remove the finished tasks from the running set. (Do this here
+                # to avoid mutating the set while iterating over it.)
+                running -= finished
+
+                # Terminate any tasks whose handlers have asked us to do so.
+                for t in terminate:
+                    t.pipe.terminate()
+                    t.pipe.wait()
+                    running.remove(t)
+
+                # Terminate any tasks which have missed their deadline.
+                finished = set()
+                for t in running:
+                    if now >= t.deadline:
+                        t.pipe.terminate()
+                        t.pipe.wait()
+                        t.onTimeout()
+                        finished.add(t)
+                # Remove the finished tasks from the running set. (Do this here
+                # to avoid mutating the set while iterating over it.)
+                running -= finished
+        return None
+
+
+def get_cpu_count():
+    """
+    Guess at a reasonable parallelism count to set as the default for the
+    current machine and run.
+    """
+    # Python 2.6+
+    try:
+        import multiprocessing
+
+        return multiprocessing.cpu_count()
+    except (ImportError, NotImplementedError):
+        pass
+
+    # POSIX
+    try:
+        res = int(os.sysconf("SC_NPROCESSORS_ONLN"))
+        if res > 0:
+            return res
+    except (AttributeError, ValueError):
+        pass
+
+    # Windows
+    try:
+        res = int(os.environ["NUMBER_OF_PROCESSORS"])
+        if res > 0:
+            return res
+    except (KeyError, ValueError):
+        pass
+
+    return 1
+
+
+if __name__ == "__main__":
+    # Test TaskPool by using it to implement the unique 'sleep sort' algorithm.
+    def sleep_sort(ns, timeout):
+        sorted = []
+
+        class SortableTask(TaskPool.Task):
+            def __init__(self, n):
+                super(SortableTask, self).__init__()
+                self.n = n
+
+            def start(self, pipe, deadline):
+                super(SortableTask, self).start(pipe, deadline)
+
+            def cmd(self):
+                return ["sh", "-c", "echo out; sleep %d; echo err>&2" % (self.n,)]
+
+            def onStdout(self, text):
+                print("%d stdout: %r" % (self.n, text))
+
+            def onStderr(self, text):
+                print("%d stderr: %r" % (self.n, text))
+
+            def onFinished(self, returncode):
+                print("%d (rc=%d)" % (self.n, returncode))
+                sorted.append(self.n)
+
+            def onTimeout(self):
+                print("%d timed out" % (self.n,))
+
+        p = TaskPool([SortableTask(_) for _ in ns], job_limit=len(ns), timeout=timeout)
+        p.run_all()
+        return sorted
+
+    print(repr(sleep_sort([1, 1, 2, 3, 5, 8, 13, 21, 34], 15)))
diff --git a/js/src/gdb/tests/enum-printers.cpp b/js/src/gdb/tests/enum-printers.cpp
new file mode 100644
index 0000000000..dfb1cd7b46
--- /dev/null
+++ b/js/src/gdb/tests/enum-printers.cpp
@@ -0,0 +1,25 @@
+#include "gdb-tests.h"
+
+#include <stdint.h>
+
+enum unscoped_no_storage { EnumValue1 };
+
+enum unscoped_with_storage : uint8_t { EnumValue2 };
+
+enum class scoped_no_storage { EnumValue3 };
+
+enum class scoped_with_storage : uint8_t { EnumValue4 };
+
+FRAGMENT(enum_printers, one) {
+  unscoped_no_storage i1 = EnumValue1;
+  unscoped_with_storage i2 = EnumValue2;
+  scoped_no_storage i3 = scoped_no_storage::EnumValue3;
+  scoped_with_storage i4 = scoped_with_storage::EnumValue4;
+
+  breakpoint();
+
+  use(i1);
+  use(i2);
+  use(i3);
+  use(i4);
+}
diff --git a/js/src/gdb/tests/enum-printers.py b/js/src/gdb/tests/enum-printers.py
new file mode 100644
index 0000000000..ff2c088dc8
--- /dev/null
+++ b/js/src/gdb/tests/enum-printers.py
@@ -0,0 +1,47 @@
+# Test that we can find pretty-printers for enums.
+# flake8: noqa: F821
+
+import mozilla.prettyprinters
+
+
+@mozilla.prettyprinters.pretty_printer("unscoped_no_storage")
+class my_typedef(object):
+    def __init__(self, value, cache):
+        pass
+
+    def to_string(self):
+        return "unscoped_no_storage::success"
+
+
+@mozilla.prettyprinters.pretty_printer("unscoped_with_storage")
+class my_typedef(object):
+    def __init__(self, value, cache):
+        pass
+
+    def to_string(self):
+        return "unscoped_with_storage::success"
+
+
+@mozilla.prettyprinters.pretty_printer("scoped_no_storage")
+class my_typedef(object):
+    def __init__(self, value, cache):
+        pass
+
+    def to_string(self):
+        return "scoped_no_storage::success"
+
+
+@mozilla.prettyprinters.pretty_printer("scoped_with_storage")
+class my_typedef(object):
+    def __init__(self, value, cache):
+        pass
+
+    def to_string(self):
+        return "scoped_with_storage::success"
+
+
+run_fragment("enum_printers.one")
+assert_pretty("i1", "unscoped_no_storage::success")
+assert_pretty("i2", "unscoped_with_storage::success")
+assert_pretty("i3", "scoped_no_storage::success")
+assert_pretty("i4", "scoped_with_storage::success")
diff --git a/js/src/gdb/tests/test-ExecutableAllocator.cpp b/js/src/gdb/tests/test-ExecutableAllocator.cpp
new file mode 100644
index 0000000000..9eeab522fe
--- /dev/null
+++ b/js/src/gdb/tests/test-ExecutableAllocator.cpp
@@ -0,0 +1,50 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+
+#include "jit/ExecutableAllocator.h"
+#include "vm/JSContext.h"
+
+FRAGMENT(ExecutableAllocator, empty) {
+  using namespace js::jit;
+  ExecutableAllocator execAlloc;
+
+  breakpoint();
+
+  use(execAlloc);
+}
+
+FRAGMENT(ExecutableAllocator, onepool) {
+  using namespace js::jit;
+  ExecutablePool* pool = nullptr;
+  ExecutableAllocator execAlloc;
+  execAlloc.alloc(cx, 16 * 1024, &pool, CodeKind::Baseline);
+
+  breakpoint();
+
+  use(pool);
+  use(execAlloc);
+}
+
+FRAGMENT(ExecutableAllocator, twopools) {
+  using namespace js::jit;
+  const size_t INIT_ALLOC_SIZE = 16 * 1024;
+  const size_t ALLOC_SIZE = 32 * 1024;
+  ExecutablePool* init = nullptr;
+  ExecutablePool* pool = nullptr;
+  ExecutableAllocator execAlloc;
+  size_t allocated = 0;
+
+  execAlloc.alloc(cx, INIT_ALLOC_SIZE, &init, CodeKind::Baseline);
+
+  do {  // Keep allocating until we get a second pool.
+    execAlloc.alloc(cx, ALLOC_SIZE, &pool, CodeKind::Ion);
+    allocated += ALLOC_SIZE;
+  } while (pool == init);
+
+  breakpoint();
+
+  use(execAlloc);
+  init->release(INIT_ALLOC_SIZE, CodeKind::Baseline);
+  init->release(allocated - ALLOC_SIZE, CodeKind::Ion);
+  pool->release(ALLOC_SIZE, CodeKind::Ion);
+}
diff --git a/js/src/gdb/tests/test-ExecutableAllocator.py b/js/src/gdb/tests/test-ExecutableAllocator.py
new file mode 100644
index 0000000000..bec2dda623
--- /dev/null
+++ b/js/src/gdb/tests/test-ExecutableAllocator.py
@@ -0,0 +1,22 @@
+# Tests for ExecutableAllocator pretty-printing
+# Ignore flake8 errors "undefined name 'assert_regexp_pretty'"
+# As it caused by the way we instanciate this file
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "JS::GCCellPtr")
+
+run_fragment("ExecutableAllocator.empty")
+
+assert_pretty("execAlloc", "ExecutableAllocator([])")
+
+run_fragment("ExecutableAllocator.onepool")
+
+reExecPool = "ExecutablePool [a-f0-9]{8,}-[a-f0-9]{8,}"
+assert_regexp_pretty("pool", reExecPool)
+assert_regexp_pretty("execAlloc", "ExecutableAllocator\(\[" + reExecPool + "\]\)")
+
+run_fragment("ExecutableAllocator.twopools")
+
+assert_regexp_pretty(
+    "execAlloc", "ExecutableAllocator\(\[" + reExecPool + ", " + reExecPool + "\]\)"
+)
diff --git a/js/src/gdb/tests/test-GCCellPtr.cpp b/js/src/gdb/tests/test-GCCellPtr.cpp
new file mode 100644
index 0000000000..87c4b13bb1
--- /dev/null
+++ b/js/src/gdb/tests/test-GCCellPtr.cpp
@@ -0,0 +1,69 @@
+#include "gdb-tests.h"
+
+#include "mozilla/Unused.h"
+
+#include "jsapi.h"
+
+#include "js/CompileOptions.h"
+#include "js/CompilationAndEvaluation.h"
+#include "js/HeapAPI.h"
+#include "js/RegExpFlags.h"
+#include "js/SourceText.h"
+#include "js/Symbol.h"
+#include "js/TypeDecls.h"
+#include "vm/BigIntType.h"
+#include "vm/JSObject.h"
+#include "vm/ObjectGroup.h"
+#include "vm/RegExpObject.h"
+#include "vm/Shape.h"
+
+#include "vm/JSObject-inl.h"
+
+FRAGMENT(GCCellPtr, simple) {
+  JS::Rooted<JSObject*> glob(cx, JS::CurrentGlobalOrNull(cx));
+  JS::Rooted<JSString*> empty(cx, JS_NewStringCopyN(cx, nullptr, 0));
+  JS::Rooted<JS::Symbol*> unique(cx, JS::NewSymbol(cx, nullptr));
+  JS::Rooted<JS::BigInt*> zeroBigInt(cx, JS::BigInt::zero(cx));
+  JS::Rooted<js::ObjectGroup*> rootedObjGroup(cx, glob->group());
+  JS::Rooted<js::RegExpObject*> regExp(
+      cx, js::RegExpObject::create(cx, u"", 0, JS::RegExpFlags{},
+                                   js::GenericObject));
+  JS::Rooted<js::RegExpShared*> rootedRegExpShared(
+      cx, js::RegExpObject::getShared(cx, regExp));
+
+  JS::CompileOptions options(cx);
+  options.setFileAndLine(__FILE__, __LINE__);
+  JS::SourceText<char16_t> srcBuf;
+  mozilla::Unused << srcBuf.init(cx, nullptr, 0, JS::SourceOwnership::Borrowed);
+  JS::RootedScript emptyScript(cx, JS::Compile(cx, options, srcBuf));
+
+  // Inline TraceKinds.
+  JS::GCCellPtr nulll(nullptr);
+  JS::GCCellPtr object(glob.get());
+  JS::GCCellPtr string(empty.get());
+  JS::GCCellPtr symbol(unique.get());
+  JS::GCCellPtr bigint(zeroBigInt.get());
+  JS::GCCellPtr shape(glob->shape());
+  JS::GCCellPtr objectGroup(rootedObjGroup.get());
+
+  // Out-of-line TraceKinds.
+  JS::GCCellPtr baseShape(glob->shape()->base());
+  // JitCode can't easily be tested here, so skip it.
+  JS::GCCellPtr script(emptyScript.get());
+  JS::GCCellPtr scope(emptyScript->bodyScope());
+  JS::GCCellPtr regExpShared(rootedRegExpShared.get());
+
+  breakpoint();
+
+  use(nulll);
+  use(object);
+  use(string);
+  use(symbol);
+  use(bigint);
+  use(shape);
+  use(objectGroup);
+  use(baseShape);
+  use(script);
+  use(scope);
+  use(regExpShared);
+}
diff --git a/js/src/gdb/tests/test-GCCellPtr.py b/js/src/gdb/tests/test-GCCellPtr.py
new file mode 100644
index 0000000000..56207c8863
--- /dev/null
+++ b/js/src/gdb/tests/test-GCCellPtr.py
@@ -0,0 +1,18 @@
+# Tests for GCCellPtr pretty-printing
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "JS::GCCellPtr")
+
+run_fragment("GCCellPtr.simple")
+
+assert_pretty("nulll", "JS::GCCellPtr(nullptr)")
+assert_pretty("object", "JS::GCCellPtr((JSObject*) )")
+assert_pretty("string", "JS::GCCellPtr((JSString*) )")
+assert_pretty("symbol", "JS::GCCellPtr((JS::Symbol*) )")
+assert_pretty("bigint", "JS::GCCellPtr((JS::BigInt*) )")
+assert_pretty("shape", "JS::GCCellPtr((js::Shape*) )")
+assert_pretty("objectGroup", "JS::GCCellPtr((js::ObjectGroup*) )")
+assert_pretty("baseShape", "JS::GCCellPtr((js::BaseShape*) )")
+assert_pretty("script", "JS::GCCellPtr((js::BaseScript*) )")
+assert_pretty("scope", "JS::GCCellPtr((js::Scope*) )")
+assert_pretty("regExpShared", "JS::GCCellPtr((js::RegExpShared*) )")
diff --git a/js/src/gdb/tests/test-Interpreter.cpp b/js/src/gdb/tests/test-Interpreter.cpp
new file mode 100644
index 0000000000..93df721010
--- /dev/null
+++ b/js/src/gdb/tests/test-Interpreter.cpp
@@ -0,0 +1,89 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
+ * vim: set ts=8 sts=2 et sw=2 tw=80:
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+#include "gdb-tests.h"
+#include "jsapi.h"
+
+#include "vm/Stack.h"
+
+namespace js {
+
+void GDBTestInitInterpreterRegs(InterpreterRegs& regs,
+                                js::InterpreterFrame* fp_, JS::Value* sp,
+                                uint8_t* pc) {
+  regs.fp_ = fp_;
+  regs.sp = sp;
+  regs.pc = pc;
+}
+
+void GDBTestInitAbstractFramePtr(AbstractFramePtr& frame,
+                                 InterpreterFrame* ptr) {
+  MOZ_ASSERT((uintptr_t(ptr) & AbstractFramePtr::TagMask) == 0);
+  frame.ptr_ = uintptr_t(ptr) | AbstractFramePtr::Tag_InterpreterFrame;
+}
+
+void GDBTestInitAbstractFramePtr(AbstractFramePtr& frame,
+                                 jit::BaselineFrame* ptr) {
+  MOZ_ASSERT((uintptr_t(ptr) & AbstractFramePtr::TagMask) == 0);
+  frame.ptr_ = uintptr_t(ptr) | AbstractFramePtr::Tag_BaselineFrame;
+}
+
+void GDBTestInitAbstractFramePtr(AbstractFramePtr& frame,
+                                 jit::RematerializedFrame* ptr) {
+  MOZ_ASSERT((uintptr_t(ptr) & AbstractFramePtr::TagMask) == 0);
+  frame.ptr_ = uintptr_t(ptr) | AbstractFramePtr::Tag_RematerializedFrame;
+}
+
+void GDBTestInitAbstractFramePtr(AbstractFramePtr& frame,
+                                 wasm::DebugFrame* ptr) {
+  MOZ_ASSERT((uintptr_t(ptr) & AbstractFramePtr::TagMask) == 0);
+  frame.ptr_ = uintptr_t(ptr) | AbstractFramePtr::Tag_WasmDebugFrame;
+}
+
+}  // namespace js
+
+FRAGMENT(Interpreter, Regs) {
+  struct FakeFrame {
+    js::InterpreterFrame frame;
+    JS::Value slot0;
+    JS::Value slot1;
+    JS::Value slot2;
+  } fakeFrame;
+  uint8_t fakeOpcode = uint8_t(JSOp::True);
+
+  js::InterpreterRegs regs;
+  js::GDBTestInitInterpreterRegs(regs, &fakeFrame.frame, &fakeFrame.slot2,
+                                 &fakeOpcode);
+
+  breakpoint();
+
+  use(regs);
+}
+
+FRAGMENT(Interpreter, AbstractFramePtr) {
+  js::AbstractFramePtr ifptr;
+  GDBTestInitAbstractFramePtr(ifptr,
+                              (js::InterpreterFrame*)uintptr_t(0x8badf00));
+
+  js::AbstractFramePtr bfptr;
+  GDBTestInitAbstractFramePtr(bfptr,
+                              (js::jit::BaselineFrame*)uintptr_t(0xbadcafe0));
+
+  js::AbstractFramePtr rfptr;
+  GDBTestInitAbstractFramePtr(
+      rfptr, (js::jit::RematerializedFrame*)uintptr_t(0xdabbad00));
+
+  js::AbstractFramePtr sfptr;
+  GDBTestInitAbstractFramePtr(sfptr,
+                              (js::wasm::DebugFrame*)uintptr_t(0xcb98ad00));
+
+  breakpoint();
+
+  use(ifptr);
+  use(bfptr);
+  use(rfptr);
+  use(sfptr);
+}
diff --git a/js/src/gdb/tests/test-Interpreter.py b/js/src/gdb/tests/test-Interpreter.py
new file mode 100644
index 0000000000..7a0279dc4d
--- /dev/null
+++ b/js/src/gdb/tests/test-Interpreter.py
@@ -0,0 +1,23 @@
+# Test printing interpreter internal data structures.
+# Ignore flake8 errors "undefined name 'assert_pretty'"
+# As it caused by the way we instanciate this file
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "js::InterpreterRegs")
+
+run_fragment("Interpreter.Regs")
+
+assert_pretty("regs", "{ fp_ = , sp = fp_.slots() + 2, pc =  (JSOp::True) }")
+
+run_fragment("Interpreter.AbstractFramePtr")
+
+assert_pretty(
+    "ifptr", "AbstractFramePtr ((js::InterpreterFrame *) ) = {ptr_ = 146464513}"
+)
+assert_pretty(
+    "bfptr", "AbstractFramePtr ((js::jit::BaselineFrame *) ) = {ptr_ = 3135025122}"
+)
+assert_pretty(
+    "rfptr",
+    "AbstractFramePtr ((js::jit::RematerializedFrame *) ) = {ptr_ = 3669732611}",
+)
diff --git a/js/src/gdb/tests/test-JSObject-null.py b/js/src/gdb/tests/test-JSObject-null.py
new file mode 100644
index 0000000000..fcda681aea
--- /dev/null
+++ b/js/src/gdb/tests/test-JSObject-null.py
@@ -0,0 +1,8 @@
+# flake8: noqa: F821
+
+gdb.execute("set print address on")
+
+run_fragment("JSObject.null")
+
+assert_pretty("null", "0x0")
+assert_pretty("nullRaw", "0x0")
diff --git a/js/src/gdb/tests/test-JSObject.cpp b/js/src/gdb/tests/test-JSObject.cpp
new file mode 100644
index 0000000000..392c0be5ba
--- /dev/null
+++ b/js/src/gdb/tests/test-JSObject.cpp
@@ -0,0 +1,52 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+#include "js/Object.h"  // JS::GetClass
+
+FRAGMENT(JSObject, simple) {
+  AutoSuppressHazardsForTest noanalysis;
+
+  JS::Rooted<JSObject*> glob(cx, JS::CurrentGlobalOrNull(cx));
+  JS::Rooted<JSObject*> plain(cx, JS_NewPlainObject(cx));
+  JS::Rooted<JSObject*> global(cx, JS::CurrentGlobalOrNull(cx));
+  JS::Rooted<JSObject*> func(
+      cx, (JSObject*)JS_NewFunction(cx, (JSNative)1, 0, 0, "dys"));
+  JS::Rooted<JSObject*> anon(
+      cx, (JSObject*)JS_NewFunction(cx, (JSNative)1, 0, 0, nullptr));
+  JS::Rooted<JSFunction*> funcPtr(
+      cx, JS_NewFunction(cx, (JSNative)1, 0, 0, "formFollows"));
+
+  JSObject& plainRef = *plain;
+  JSFunction& funcRef = *funcPtr;
+  JSObject* plainRaw = plain;
+  JSObject* funcRaw = func;
+
+  // JS_NewObject will now assert if you feed it a bad class name, so mangle
+  // the name after construction.
+  char namebuf[20] = "goodname";
+  static JSClass cls{namebuf};
+  JS::RootedObject badClassName(cx, JS_NewObject(cx, &cls));
+  strcpy(namebuf, "\xc7X");
+
+  breakpoint();
+
+  use(glob);
+  use(plain);
+  use(func);
+  use(anon);
+  use(funcPtr);
+  use(&plainRef);
+  use(&funcRef);
+  use(JS::GetClass((JSObject*)&funcRef));
+  use(plainRaw);
+  use(funcRaw);
+}
+
+FRAGMENT(JSObject, null) {
+  JS::Rooted<JSObject*> null(cx, nullptr);
+  JSObject* nullRaw = null;
+
+  breakpoint();
+
+  use(null);
+  use(nullRaw);
+}
diff --git a/js/src/gdb/tests/test-JSObject.py b/js/src/gdb/tests/test-JSObject.py
new file mode 100644
index 0000000000..a9a5ca41a6
--- /dev/null
+++ b/js/src/gdb/tests/test-JSObject.py
@@ -0,0 +1,23 @@
+# Printing JSObjects.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "ptr-to-JSObject")
+assert_subprinter_registered("SpiderMonkey", "ref-to-JSObject")
+
+run_fragment("JSObject.simple")
+
+# These patterns look a little strange because of prologue.py's 'set print
+# address off', which avoids putting varying addresses in the output. After
+# the '(JSObject *) ', there is a 'void *' value printing as the empty
+# string.
+
+assert_pretty("glob", "(JSObject *)  [object global] delegate")
+assert_pretty("plain", "(JSObject *)  [object Object]")
+assert_pretty("func", '(JSObject *)  [object Function "dys"]')
+assert_pretty("anon", "(JSObject *)  [object Function <unnamed>]")
+assert_pretty("funcPtr", '(JSFunction *)  [object Function "formFollows"]')
+
+assert_pretty("badClassName", "(JSObject *)  [object \\307X]")
+
+assert_pretty("plainRef", "(JSObject &) @ [object Object]")
+assert_pretty("funcRef", '(JSFunction &) @ [object Function "formFollows"]')
diff --git a/js/src/gdb/tests/test-JSString-null.py b/js/src/gdb/tests/test-JSString-null.py
new file mode 100644
index 0000000000..26bbfcf428
--- /dev/null
+++ b/js/src/gdb/tests/test-JSString-null.py
@@ -0,0 +1,8 @@
+# flake8: noqa: F821
+
+gdb.execute("set print address on")
+
+run_fragment("JSString.null")
+
+assert_pretty("null", "0x0")
+assert_pretty("nullRaw", "0x0")
diff --git a/js/src/gdb/tests/test-JSString-subclasses.py b/js/src/gdb/tests/test-JSString-subclasses.py
new file mode 100644
index 0000000000..b685f46aff
--- /dev/null
+++ b/js/src/gdb/tests/test-JSString-subclasses.py
@@ -0,0 +1,7 @@
+# flake8: noqa: F821
+
+# We can print pointers to subclasses of JSString.
+
+run_fragment("JSString.subclasses")
+
+assert_pretty("linear", '"Hi!"')
diff --git a/js/src/gdb/tests/test-JSString.cpp b/js/src/gdb/tests/test-JSString.cpp
new file mode 100644
index 0000000000..8d3dfc6ea5
--- /dev/null
+++ b/js/src/gdb/tests/test-JSString.cpp
@@ -0,0 +1,69 @@
+#include "gdb-tests.h"
+
+#include "vm/JSContext.h"
+// When JSGC_ANALYSIS is #defined, Rooted<JSLinearString*> needs the definition
+// of JSLinearString in order to figure out its ThingRootKind
+#include "vm/StringType.h"
+
+FRAGMENT(JSString, simple) {
+  AutoSuppressHazardsForTest noanalysis;
+
+  JS::Rooted<JSString*> empty(cx, JS_NewStringCopyN(cx, nullptr, 0));
+  JS::Rooted<JSString*> x(cx, JS_NewStringCopyN(cx, "x", 1));
+  JS::Rooted<JSString*> z(cx, JS_NewStringCopyZ(cx, "z"));
+
+  // I expect this will be a non-inlined string.
+  JS::Rooted<JSString*> stars(cx,
+                              JS_NewStringCopyZ(cx,
+                                                "*************************"
+                                                "*************************"
+                                                "*************************"
+                                                "*************************"));
+
+  // This may well be an inlined string.
+  JS::Rooted<JSString*> xz(cx, JS_ConcatStrings(cx, x, z));
+
+  // This will probably be a rope.
+  JS::Rooted<JSString*> doubleStars(cx, JS_ConcatStrings(cx, stars, stars));
+
+  // Ensure we're not confused by typedefs for pointer types.
+  JSString* xRaw = x;
+
+  breakpoint();
+
+  use(empty);
+  use(x);
+  use(z);
+  use(stars);
+  use(xz);
+  use(doubleStars);
+  use(xRaw);
+}
+
+FRAGMENT(JSString, null) {
+  AutoSuppressHazardsForTest noanalysis;
+
+  JS::Rooted<JSString*> null(cx, nullptr);
+  JSString* nullRaw = null;
+
+  breakpoint();
+
+  use(null);
+  use(nullRaw);
+}
+
+FRAGMENT(JSString, subclasses) {
+  JS::Rooted<JSLinearString*> linear(
+      cx, JS_EnsureLinearString(cx, JS_NewStringCopyZ(cx, "Hi!")));
+
+  breakpoint();
+
+  use(linear);
+}
+
+FRAGMENT(JSString, atom) {
+  JSAtom* molybdenum = js::Atomize(cx, "molybdenum", 10);
+  breakpoint();
+
+  use(molybdenum);
+}
diff --git a/js/src/gdb/tests/test-JSString.py b/js/src/gdb/tests/test-JSString.py
new file mode 100644
index 0000000000..99cebc4ab5
--- /dev/null
+++ b/js/src/gdb/tests/test-JSString.py
@@ -0,0 +1,24 @@
+# Printing JSStrings.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "ptr-to-JSString")
+run_fragment("JSString.simple")
+
+assert_pretty("empty", '""')
+assert_pretty("x", '"x"')
+assert_pretty("z", '"z"')
+assert_pretty("xz", '"xz"')
+
+stars = gdb.parse_and_eval("stars")
+assert_eq(str(stars), "'*' <repeats 100 times>")
+
+doubleStars = gdb.parse_and_eval("doubleStars")
+assert_eq(str(doubleStars), "'*' <repeats 200 times>")
+
+assert_pretty("xRaw", '"x"')
+
+# JSAtom *
+
+run_fragment("JSString.atom")
+
+assert_pretty("molybdenum", '"molybdenum"')
diff --git a/js/src/gdb/tests/test-JSSymbol.cpp b/js/src/gdb/tests/test-JSSymbol.cpp
new file mode 100644
index 0000000000..29695731e3
--- /dev/null
+++ b/js/src/gdb/tests/test-JSSymbol.cpp
@@ -0,0 +1,22 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+
+#include "js/Symbol.h"
+
+FRAGMENT(JSSymbol, simple) {
+  using namespace JS;
+
+  RootedString hello(cx, JS_NewStringCopyZ(cx, "Hello!"));
+
+  Rooted<Symbol*> unique(cx, NewSymbol(cx, nullptr));
+  Rooted<Symbol*> unique_with_desc(cx, NewSymbol(cx, hello));
+  Rooted<Symbol*> registry(cx, GetSymbolFor(cx, hello));
+  Rooted<Symbol*> well_known(cx, GetWellKnownSymbol(cx, SymbolCode::iterator));
+
+  breakpoint();
+
+  use(unique);
+  use(unique_with_desc);
+  use(registry);
+  use(well_known);
+}
diff --git a/js/src/gdb/tests/test-JSSymbol.py b/js/src/gdb/tests/test-JSSymbol.py
new file mode 100644
index 0000000000..a0dba5c036
--- /dev/null
+++ b/js/src/gdb/tests/test-JSSymbol.py
@@ -0,0 +1,11 @@
+# Printing JS::Symbols.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "ptr-to-JS::Symbol")
+
+run_fragment("JSSymbol.simple")
+
+assert_pretty("unique", "Symbol()")
+assert_pretty("unique_with_desc", 'Symbol("Hello!")')
+assert_pretty("registry", 'Symbol.for("Hello!")')
+assert_pretty("well_known", "Symbol.iterator")
diff --git a/js/src/gdb/tests/test-Root-null.py b/js/src/gdb/tests/test-Root-null.py
new file mode 100644
index 0000000000..839fa069c5
--- /dev/null
+++ b/js/src/gdb/tests/test-Root-null.py
@@ -0,0 +1,21 @@
+# Test printing roots that refer to NULL pointers.
+
+# Since mozilla.prettyprinters.Pointer declines to create pretty-printers
+# for null pointers, GDB built-in printing code ends up handling them. But
+# as of 2012-11, GDB suppresses printing pointers in replacement values:
+# see: http://sourceware.org/ml/gdb/2012-11/msg00055.html
+#
+# Thus, if the pretty-printer for JS::Rooted simply returns the referent as
+# a replacement value (which seems reasonable enough, if you want the
+# pretty-printer to be completely transparent), and the referent is a null
+# pointer, it prints as nothing at all.
+#
+# This test ensures that the JS::Rooted pretty-printer doesn't make that
+# mistake.
+# flake8: noqa: F821
+
+gdb.execute("set print address on")
+
+run_fragment("Root.null")
+
+assert_pretty("null", "0x0")
diff --git a/js/src/gdb/tests/test-Root.cpp b/js/src/gdb/tests/test-Root.cpp
new file mode 100644
index 0000000000..2d93de469d
--- /dev/null
+++ b/js/src/gdb/tests/test-Root.cpp
@@ -0,0 +1,63 @@
+#include "gdb-tests.h"
+
+#include "jsapi.h"
+
+#include "gc/Barrier.h"
+#include "js/Array.h"  // JS::NewArrayObject
+#include "vm/JSFunction.h"
+
+FRAGMENT(Root, null) {
+  JS::Rooted<JSObject*> null(cx, nullptr);
+
+  breakpoint();
+
+  use(null);
+}
+
+void callee(JS::Handle<JSObject*> obj,
+            JS::MutableHandle<JSObject*> mutableObj) {
+  // Prevent the linker from unifying this function with others that are
+  // equivalent in machine code but not type.
+  fprintf(stderr, "Called " __FILE__ ":callee\n");
+  breakpoint();
+}
+
+FRAGMENT(Root, handle) {
+  JS::Rooted<JSObject*> global(cx, JS::CurrentGlobalOrNull(cx));
+  callee(global, &global);
+  use(global);
+}
+
+FRAGMENT(Root, HeapSlot) {
+  JS::Rooted<JS::Value> plinth(
+      cx, JS::StringValue(JS_NewStringCopyZ(cx, "plinth")));
+  JS::Rooted<JSObject*> array(
+      cx, JS::NewArrayObject(cx, JS::HandleValueArray(plinth)));
+
+  breakpoint();
+
+  use(plinth);
+  use(array);
+}
+
+FRAGMENT(Root, barriers) {
+  JSObject* obj = JS_NewPlainObject(cx);
+  js::PreBarriered<JSObject*> prebarriered(obj);
+  js::GCPtrObject heapptr(obj);
+  js::HeapPtr<JSObject*> relocatable(obj);
+
+  JS::Value val = JS::ObjectValue(*obj);
+  js::PreBarrieredValue prebarrieredValue(JS::ObjectValue(*obj));
+  js::GCPtrValue heapValue(JS::ObjectValue(*obj));
+  js::HeapPtr<JS::Value> relocatableValue(JS::ObjectValue(*obj));
+
+  breakpoint();
+
+  use(prebarriered);
+  use(heapptr);
+  use(relocatable);
+  use(val);
+  use(prebarrieredValue);
+  use(heapValue);
+  use(relocatableValue);
+}
diff --git a/js/src/gdb/tests/test-Root.py b/js/src/gdb/tests/test-Root.py
new file mode 100644
index 0000000000..f3b72b9fda
--- /dev/null
+++ b/js/src/gdb/tests/test-Root.py
@@ -0,0 +1,30 @@
+# Test printing Handles.
+# Ignore flake8 errors "undefined name 'assert_pretty'"
+# As it caused by the way we instanciate this file
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "instantiations-of-JS::Rooted")
+assert_subprinter_registered("SpiderMonkey", "instantiations-of-JS::Handle")
+assert_subprinter_registered("SpiderMonkey", "instantiations-of-JS::MutableHandle")
+assert_subprinter_registered("SpiderMonkey", "instantiations-of-js::BarrieredBase")
+
+run_fragment("Root.handle")
+
+assert_pretty("obj", "(JSObject * const)  [object global] delegate")
+assert_pretty("mutableObj", "(JSObject *)  [object global] delegate")
+
+run_fragment("Root.HeapSlot")
+
+# This depends on implementation details of arrays, but since HeapSlot is
+# not a public type, I'm not sure how to avoid doing *something* ugly.
+assert_pretty("((js::NativeObject *) array.ptr)->elements_[0]", '$JS::Value("plinth")')
+
+run_fragment("Root.barriers")
+
+assert_pretty("prebarriered", "(JSObject *)  [object Object]")
+assert_pretty("heapptr", "(JSObject *)  [object Object]")
+assert_pretty("relocatable", "(JSObject *)  [object Object]")
+assert_pretty("val", "$JS::Value((JSObject *)  [object Object])")
+assert_pretty("heapValue", "$JS::Value((JSObject *)  [object Object])")
+assert_pretty("prebarrieredValue", "$JS::Value((JSObject *)  [object Object])")
+assert_pretty("relocatableValue", "$JS::Value((JSObject *)  [object Object])")
diff --git a/js/src/gdb/tests/test-asmjs.cpp b/js/src/gdb/tests/test-asmjs.cpp
new file mode 100644
index 0000000000..e58b25db74
--- /dev/null
+++ b/js/src/gdb/tests/test-asmjs.cpp
@@ -0,0 +1,45 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+#include "js/CompilationAndEvaluation.h"
+#include "js/CompileOptions.h"
+#include "js/RootingAPI.h"
+#include "js/SourceText.h"
+#include "js/Value.h"
+#include "mozilla/Utf8.h"
+#include "util/Text.h"
+
+#include <string.h>
+
+FRAGMENT(asmjs, segfault) {
+  constexpr unsigned line0 = __LINE__;
+  static const char chars[] =
+      "function f(glob, ffi, heap) {\n"
+      "    \"use asm\";\n"
+      "    var f32 = new glob.Float32Array(heap);\n"
+      "    function g(n) {\n"
+      "        n = n | 0;\n"
+      "        return +f32[n>>2];\n"
+      "    }\n"
+      "    return g;\n"
+      "}\n"
+      "\n"
+      "var func = f(this, null, new ArrayBuffer(0x10000));\n"
+      "func(0x10000 << 2);\n"
+      "'ok'\n";
+
+  JS::CompileOptions opts(cx);
+  opts.setFileAndLine(__FILE__, line0 + 1);
+  opts.asmJSOption = JS::AsmJSOption::Enabled;
+
+  JS::SourceText<mozilla::Utf8Unit> srcBuf;
+  JS::Rooted<JS::Value> rval(cx);
+
+  bool ok =
+      srcBuf.init(cx, chars, js_strlen(chars), JS::SourceOwnership::Borrowed) &&
+      JS::Evaluate(cx, opts, srcBuf, &rval);
+
+  breakpoint();
+
+  use(ok);
+  use(rval);
+}
diff --git a/js/src/gdb/tests/test-asmjs.py b/js/src/gdb/tests/test-asmjs.py
new file mode 100644
index 0000000000..9e3ed7fc89
--- /dev/null
+++ b/js/src/gdb/tests/test-asmjs.py
@@ -0,0 +1,16 @@
+# Test for special asmjs SIGSEGV-handling.
+#
+# Expected behavior is for the asm.js code in the following fragment to trigger
+# SIGSEGV. The code in js/src/gdb/mozilla/asmjs.py should prevent GDB from
+# handling that signal.
+# flake8: noqa: F821
+
+run_fragment("asmjs.segfault")
+
+# If SIGSEGV handling is broken, GDB would have stopped at the SIGSEGV signal.
+# The breakpoint would not have hit, and run_fragment would have thrown.
+#
+# So if we get here, and the asm.js code actually ran, we win.
+
+assert_pretty("ok", "true")
+assert_pretty("rval", '$JS::Value("ok")')
diff --git a/js/src/gdb/tests/test-jsbytecode.cpp b/js/src/gdb/tests/test-jsbytecode.cpp
new file mode 100644
index 0000000000..6747b488d1
--- /dev/null
+++ b/js/src/gdb/tests/test-jsbytecode.cpp
@@ -0,0 +1,33 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+#include "js/CompilationAndEvaluation.h"
+#include "js/CompileOptions.h"
+#include "js/SourceText.h"
+#include "util/Text.h"
+#include "vm/JSFunction.h"
+#include "vm/JSScript.h"
+#include "mozilla/Utf8.h"
+
+FRAGMENT(jsbytecode, simple) {
+  constexpr unsigned line0 = __LINE__;
+  static const char chars[] = R"(
+    debugger;
+  )";
+
+  JS::CompileOptions opts(cx);
+  opts.setFileAndLine(__FILE__, line0 + 1);
+
+  JS::SourceText<mozilla::Utf8Unit> srcBuf;
+  JS::Rooted<JS::Value> rval(cx);
+
+  bool ok =
+      srcBuf.init(cx, chars, js_strlen(chars), JS::SourceOwnership::Borrowed);
+
+  JSScript* script = JS::Compile(cx, opts, srcBuf);
+  jsbytecode* code = script->code();
+
+  breakpoint();
+
+  use(ok);
+  use(code);
+}
diff --git a/js/src/gdb/tests/test-jsbytecode.py b/js/src/gdb/tests/test-jsbytecode.py
new file mode 100644
index 0000000000..e69732d4c7
--- /dev/null
+++ b/js/src/gdb/tests/test-jsbytecode.py
@@ -0,0 +1,9 @@
+# Basic unit tests for jsbytecode* pretty-printer.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "ptr-to-jsbytecode")
+
+run_fragment("jsbytecode.simple")
+
+assert_pretty("ok", "true")
+assert_pretty("code", " (JSOp::Debugger)")
diff --git a/js/src/gdb/tests/test-jsid.cpp b/js/src/gdb/tests/test-jsid.cpp
new file mode 100644
index 0000000000..7245b05c17
--- /dev/null
+++ b/js/src/gdb/tests/test-jsid.cpp
@@ -0,0 +1,44 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+
+#include "js/Symbol.h"
+
+FRAGMENT(jsid, simple) {
+  JS::Rooted<JSString*> string(cx, JS_NewStringCopyZ(cx, "moon"));
+  JS::Rooted<JSString*> interned(cx, JS_AtomizeAndPinJSString(cx, string));
+  JS::Rooted<jsid> string_id(cx, JS::PropertyKey::fromPinnedString(interned));
+  JS::Rooted<jsid> int_id(cx, INT_TO_JSID(1729));
+  JS::Rooted<jsid> unique_symbol_id(
+      cx, SYMBOL_TO_JSID(JS::NewSymbol(cx, interned)));
+  JS::Rooted<jsid> registry_symbol_id(
+      cx, SYMBOL_TO_JSID(JS::GetSymbolFor(cx, interned)));
+  JS::Rooted<jsid> well_known_symbol_id(
+      cx, SYMBOL_TO_JSID(JS::GetWellKnownSymbol(cx, JS::SymbolCode::iterator)));
+  jsid void_id = JSID_VOID;
+  jsid empty_id = JSID_EMPTY;
+
+  breakpoint();
+
+  use(string_id);
+  use(int_id);
+  use(unique_symbol_id);
+  use(registry_symbol_id);
+  use(well_known_symbol_id);
+  use(void_id);
+  use(empty_id);
+}
+
+void jsid_handles(JS::Handle<jsid> jsid_handle,
+                  JS::MutableHandle<jsid> mutable_jsid_handle) {
+  // Prevent the linker from unifying this function with others that are
+  // equivalent in machine code but not type.
+  fprintf(stderr, "Called " __FILE__ ":jsid_handles\n");
+  breakpoint();
+}
+
+FRAGMENT(jsid, handles) {
+  JS::Rooted<JSString*> string(cx, JS_NewStringCopyZ(cx, "shovel"));
+  JS::Rooted<JSString*> interned(cx, JS_AtomizeAndPinJSString(cx, string));
+  JS::Rooted<jsid> string_id(cx, JS::PropertyKey::fromPinnedString(interned));
+  jsid_handles(string_id, &string_id);
+}
diff --git a/js/src/gdb/tests/test-jsid.py b/js/src/gdb/tests/test-jsid.py
new file mode 100644
index 0000000000..7945c12f4c
--- /dev/null
+++ b/js/src/gdb/tests/test-jsid.py
@@ -0,0 +1,20 @@
+# Tests for jsid pretty-printing
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "JS::PropertyKey")
+
+run_fragment("jsid.simple")
+
+assert_pretty("string_id", '$jsid("moon")')
+assert_pretty("int_id", "$jsid(1729)")
+unique_symbol_pretty = str(gdb.parse_and_eval("unique_symbol_id")).split("@")[0]
+assert_eq(unique_symbol_pretty, '$jsid(Symbol("moon"))')
+assert_pretty("registry_symbol_id", '$jsid(Symbol.for("moon"))')
+assert_pretty("well_known_symbol_id", "$jsid(Symbol.iterator)")
+assert_pretty("void_id", "JSID_VOID")
+assert_pretty("empty_id", "JSID_EMPTY")
+
+run_fragment("jsid.handles")
+
+assert_pretty("jsid_handle", '$jsid("shovel")')
+assert_pretty("mutable_jsid_handle", '$jsid("shovel")')
diff --git a/js/src/gdb/tests/test-jsop.cpp b/js/src/gdb/tests/test-jsop.cpp
new file mode 100644
index 0000000000..d92c8f0d2e
--- /dev/null
+++ b/js/src/gdb/tests/test-jsop.cpp
@@ -0,0 +1,13 @@
+#include "gdb-tests.h"
+
+#include "vm/BytecodeUtil.h"
+
+FRAGMENT(jsop, simple) {
+  JSOp undefined = JSOp::Undefined;
+  JSOp debugger = JSOp::Debugger;
+
+  breakpoint();
+
+  use(undefined);
+  use(debugger);
+}
diff --git a/js/src/gdb/tests/test-jsop.py b/js/src/gdb/tests/test-jsop.py
new file mode 100644
index 0000000000..fea0da6059
--- /dev/null
+++ b/js/src/gdb/tests/test-jsop.py
@@ -0,0 +1,9 @@
+# Basic unit tests for JSOp pretty-printer.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "JSOp")
+
+run_fragment("jsop.simple")
+
+assert_pretty("undefined", "JSOp::Undefined")
+assert_pretty("debugger", "JSOp::Debugger")
diff --git a/js/src/gdb/tests/test-jsval.cpp b/js/src/gdb/tests/test-jsval.cpp
new file mode 100644
index 0000000000..9fe172e6fc
--- /dev/null
+++ b/js/src/gdb/tests/test-jsval.cpp
@@ -0,0 +1,48 @@
+#include "gdb-tests.h"
+#include "jsapi.h"
+
+#include "js/Symbol.h"
+#include "vm/BigIntType.h"
+
+FRAGMENT(jsval, simple) {
+  using namespace JS;
+
+  RootedValue fortytwo(cx, Int32Value(42));
+  RootedValue fortytwoD(cx, DoubleValue(42));
+  RootedValue negone(cx, Int32Value(-1));
+  RootedValue undefined(cx, UndefinedValue());
+  RootedValue null(cx, NullValue());
+  RootedValue js_true(cx, BooleanValue(true));
+  RootedValue js_false(cx, BooleanValue(false));
+  RootedValue elements_hole(cx, js::MagicValue(JS_ELEMENTS_HOLE));
+
+  RootedValue empty_string(cx);
+  empty_string.setString(JS_NewStringCopyZ(cx, ""));
+  RootedString hello(cx, JS_NewStringCopyZ(cx, "Hello!"));
+  RootedValue friendly_string(cx, StringValue(hello));
+  RootedValue symbol(cx, SymbolValue(GetSymbolFor(cx, hello)));
+  RootedValue bi(cx, BigIntValue(BigInt::zero(cx)));
+
+  RootedValue global(cx);
+  global.setObject(*CurrentGlobalOrNull(cx));
+
+  // Some interesting value that floating-point won't munge.
+  RootedValue onehundredthirtysevenonehundredtwentyeighths(
+      cx, DoubleValue(137.0 / 128.0));
+
+  breakpoint();
+
+  use(fortytwo);
+  use(fortytwoD);
+  use(negone);
+  use(undefined);
+  use(js_true);
+  use(js_false);
+  use(null);
+  use(elements_hole);
+  use(empty_string);
+  use(friendly_string);
+  use(symbol);
+  use(bi);
+  use(global);
+}
diff --git a/js/src/gdb/tests/test-jsval.py b/js/src/gdb/tests/test-jsval.py
new file mode 100644
index 0000000000..a7befa917c
--- /dev/null
+++ b/js/src/gdb/tests/test-jsval.py
@@ -0,0 +1,23 @@
+# Basic unit tests for jsval pretty-printer.
+# flake8: noqa: F821
+
+assert_subprinter_registered("SpiderMonkey", "JS::Value")
+
+run_fragment("jsval.simple")
+
+assert_pretty("fortytwo", "$JS::Int32Value(42)")
+assert_pretty("fortytwoD", "$JS::DoubleValue(42.0)")
+assert_pretty("negone", "$JS::Int32Value(-1)")
+assert_pretty("undefined", "$JS::UndefinedValue()")
+assert_pretty("null", "$JS::NullValue()")
+assert_pretty("js_true", "$JS::BooleanValue(true)")
+assert_pretty("js_false", "$JS::BooleanValue(false)")
+assert_pretty("elements_hole", "$JS::MagicValue(JS_ELEMENTS_HOLE)")
+assert_pretty("empty_string", '$JS::Value("")')
+assert_pretty("friendly_string", '$JS::Value("Hello!")')
+assert_pretty("symbol", '$JS::Value(Symbol.for("Hello!"))')
+assert_pretty("bi", "$JS::BigIntValue()")
+assert_pretty("global", "$JS::Value((JSObject *)  [object global] delegate)")
+assert_pretty(
+    "onehundredthirtysevenonehundredtwentyeighths", "$JS::DoubleValue(1.0703125)"
+)
diff --git a/js/src/gdb/tests/test-prettyprinters.cpp b/js/src/gdb/tests/test-prettyprinters.cpp
new file mode 100644
index 0000000000..af71c52fa3
--- /dev/null
+++ b/js/src/gdb/tests/test-prettyprinters.cpp
@@ -0,0 +1,38 @@
+#include "gdb-tests.h"
+
+typedef int A;
+typedef A B;
+
+class C {};
+class D {};
+typedef C C_;
+typedef D D_;
+class E : C, D {};
+typedef E E_;
+class F : C_, D_ {};
+class G {};
+class H : F, G {};
+
+FRAGMENT(prettyprinters, implemented_types) {
+  int i = 0;
+  A a = 0;
+  B b = 0;
+  C c;
+  C_ c_;
+  E e;
+  E_ e_;
+  F f;
+  H h;
+
+  breakpoint();
+
+  use(i);
+  use(a);
+  use(b);
+  use(c);
+  use(c_);
+  use(e);
+  use(e_);
+  use(f);
+  use(h);
+}
diff --git a/js/src/gdb/tests/test-prettyprinters.py b/js/src/gdb/tests/test-prettyprinters.py
new file mode 100644
index 0000000000..f3329c0288
--- /dev/null
+++ b/js/src/gdb/tests/test-prettyprinters.py
@@ -0,0 +1,40 @@
+# Ignore flake8 errors "undefined name 'assert_pretty'"
+# As it caused by the way we instanciate this file
+# flake8: noqa: F821
+
+import mozilla.prettyprinters
+
+run_fragment("prettyprinters.implemented_types")
+
+
+def implemented_type_names(expr):
+    v = gdb.parse_and_eval(expr)
+    it = mozilla.prettyprinters.implemented_types(v.type)
+    return [str(_) for _ in it]
+
+
+assert_eq(implemented_type_names("i"), ["int"])
+assert_eq(implemented_type_names("a"), ["A", "int"])
+assert_eq(implemented_type_names("b"), ["B", "A", "int"])
+assert_eq(implemented_type_names("c"), ["C"])
+assert_eq(implemented_type_names("c_"), ["C_", "C"])
+assert_eq(implemented_type_names("e"), ["E", "C", "D"])
+assert_eq(implemented_type_names("e_"), ["E_", "E", "C", "D"])
+
+# Some compilers strip trivial typedefs in the debuginfo from classes' base
+# classes. Sometimes this can be fixed with -fno-eliminate-unused-debug-types,
+# but not always. Allow this test to pass if the typedefs are stripped.
+#
+# It would probably be better to figure out how to make the compiler emit them,
+# since I think this test is here for a reason.
+if gdb.lookup_type("F").fields()[0].name == "C_":
+    # We have the typedef info.
+    assert_eq(implemented_type_names("f"), ["F", "C_", "D_", "C", "D"])
+    assert_eq(implemented_type_names("h"), ["H", "F", "G", "C_", "D_", "C", "D"])
+else:
+    assert_eq(implemented_type_names("f"), ["F", "C", "D"])
+    assert_eq(implemented_type_names("h"), ["H", "F", "G", "C", "D"])
+
+# Check that our pretty-printers aren't interfering with printing other types.
+assert_pretty("10", "10")
+assert_pretty("(void*) 0", "")  # Because of 'set print address off'
diff --git a/js/src/gdb/tests/test-unwind.cpp b/js/src/gdb/tests/test-unwind.cpp
new file mode 100644
index 0000000000..13a82bf790
--- /dev/null
+++ b/js/src/gdb/tests/test-unwind.cpp
@@ -0,0 +1,74 @@
+#include "gdb-tests.h"
+#include "jsapi.h"  // sundry symbols not moved to more-specific headers yet
+#include "jsfriendapi.h"  // JSFunctionSpecWithHelp
+
+#include "jit/JitOptions.h"               // js::jit::JitOptions
+#include "js/CallArgs.h"                  // JS::CallArgs, JS::CallArgsFromVp
+#include "js/CompilationAndEvaluation.h"  // JS::Evaluate
+#include "js/CompileOptions.h"            // JS::CompileOptions
+#include "js/RootingAPI.h"                // JS::Rooted
+#include "js/SourceText.h"                // JS::Source{Ownership,Text}
+#include "js/Value.h"                     // JS::Value
+
+#include "mozilla/Utf8.h"  // mozilla::Utf8Unit
+
+#include <stdint.h>  // uint32_t
+#include <string.h>  // strlen
+
+static bool Something(JSContext* cx, unsigned argc, JS::Value* vp) {
+  JS::CallArgs args = JS::CallArgsFromVp(argc, vp);
+  args.rval().setInt32(23);
+  breakpoint();
+  return true;
+}
+
+// clang-format off
+static const JSFunctionSpecWithHelp unwind_functions[] = {
+    JS_FN_HELP("something", Something, 0, 0,
+"something()",
+"  Test function for test-unwind."),
+    JS_FS_HELP_END
+};
+// clang-format on
+
+FRAGMENT(unwind, simple) {
+  using namespace JS;
+
+  JS::Rooted<JSObject*> global(cx, JS::CurrentGlobalOrNull(cx));
+  if (!JS_DefineFunctionsWithHelp(cx, global, unwind_functions)) {
+    return;
+  }
+
+  // Define an itercount property and use it to ensure Baseline compilation.
+  uint32_t threshold = js::jit::JitOptions.baselineJitWarmUpThreshold;
+  RootedValue val(cx, Int32Value(threshold + 10));
+  if (!JS_DefineProperty(cx, global, "itercount", val, 0)) {
+    return;
+  }
+
+  int line0 = __LINE__;
+  const char* bytes =
+      "\n"
+      "function unwindFunctionInner() {\n"
+      "    for (var i = 0; i < itercount; i++) {}\n"
+      "    return something();\n"
+      "}\n"
+      "\n"
+      "function unwindFunctionOuter() {\n"
+      "    for (var i = 0; i < itercount; i++) {}\n"
+      "    return unwindFunctionInner();\n"
+      "}\n"
+      "\n"
+      "unwindFunctionOuter();\n";
+
+  JS::CompileOptions opts(cx);
+  opts.setFileAndLine(__FILE__, line0 + 1);
+
+  JS::SourceText<mozilla::Utf8Unit> srcBuf;
+  if (!srcBuf.init(cx, bytes, strlen(bytes), JS::SourceOwnership::Borrowed)) {
+    return;
+  }
+
+  JS::Rooted<JS::Value> rval(cx);
+  JS::Evaluate(cx, opts, srcBuf, &rval);
+}
diff --git a/js/src/gdb/tests/test-unwind.py b/js/src/gdb/tests/test-unwind.py
new file mode 100644
index 0000000000..d12c4b0f6e
--- /dev/null
+++ b/js/src/gdb/tests/test-unwind.py
@@ -0,0 +1,61 @@
+# Test the unwinder and the frame filter.
+# flake8:  NOQA: F821
+import platform
+
+
+def do_unwinder_test():
+    # The unwinder is disabled by default for the moment. Turn it on to check
+    # that the unwinder works as expected.
+    import gdb
+
+    gdb.execute("enable unwinder .* SpiderMonkey")
+
+    run_fragment("unwind.simple", "Something")
+
+    first = True
+    # The unwinder is a bit flaky still but should at least be able to
+    # recognize one set of entry and exit frames.  This also tests to
+    # make sure we didn't end up solely in the interpreter.
+    found_entry = False
+    found_exit = False
+    found_main = False
+    found_inner = False
+    found_outer = False
+    frames = list(gdb.frames.execute_frame_filters(gdb.newest_frame(), 0, -1))
+    for frame in frames:
+        print("examining " + frame.function())
+        if first:
+            assert_eq(frame.function().startswith("Something"), True)
+            first = False
+        elif frame.function() == "<<FrameType::Exit>>":
+            found_exit = True
+        elif frame.function() == "<<FrameType::CppToJSJit>>":
+            found_entry = True
+        elif frame.function() == "main":
+            found_main = True
+        elif "unwindFunctionInner" in frame.function():
+            found_inner = True
+        elif "unwindFunctionOuter" in frame.function():
+            found_outer = True
+
+    # Had to have found a frame.
+    assert_eq(first, False)
+    # Had to have found main.
+    assert_eq(found_main, True)
+    # Had to have found the entry and exit frames.
+    assert_eq(found_exit, True)
+    assert_eq(found_entry, True)
+    # Had to have found the names of the two JS functions.
+    assert_eq(found_inner, True)
+    assert_eq(found_outer, True)
+
+
+# Only on the right platforms.
+if platform.machine() == "x86_64" and platform.system() == "Linux":
+    # Only test when gdb has the unwinder feature.
+    try:
+        import gdb.unwinder  # NOQA: F401
+
+        do_unwinder_test()
+    except Exception:
+        pass
diff --git a/js/src/gdb/tests/typedef-printers.cpp b/js/src/gdb/tests/typedef-printers.cpp
new file mode 100644
index 0000000000..efb01f82cf
--- /dev/null
+++ b/js/src/gdb/tests/typedef-printers.cpp
@@ -0,0 +1,11 @@
+#include "gdb-tests.h"
+
+typedef int my_typedef;
+
+FRAGMENT(typedef_printers, one) {
+  my_typedef i = 0;
+
+  breakpoint();
+
+  use(i);
+}
diff --git a/js/src/gdb/tests/typedef-printers.py b/js/src/gdb/tests/typedef-printers.py
new file mode 100644
index 0000000000..888379ee63
--- /dev/null
+++ b/js/src/gdb/tests/typedef-printers.py
@@ -0,0 +1,18 @@
+# Test that we can find pretty-printers for typedef names, not just for
+# struct types and templates.
+# flake8: noqa: F821
+
+import mozilla.prettyprinters
+
+
+@mozilla.prettyprinters.pretty_printer("my_typedef")
+class my_typedef(object):
+    def __init__(self, value, cache):
+        pass
+
+    def to_string(self):
+        return "huzzah"
+
+
+run_fragment("typedef_printers.one")
+assert_pretty("i", "huzzah")