diff options
Diffstat (limited to 'src/boost/tools/build/test/test_all.py')
| -rw-r--r-- | src/boost/tools/build/test/test_all.py | 373 |
1 files changed, 373 insertions, 0 deletions
diff --git a/src/boost/tools/build/test/test_all.py b/src/boost/tools/build/test/test_all.py new file mode 100644 index 000000000..651034f79 --- /dev/null +++ b/src/boost/tools/build/test/test_all.py @@ -0,0 +1,373 @@ +#!/usr/bin/python + +# Copyright 2002-2005 Dave Abrahams. +# Copyright 2002-2006 Vladimir Prus. +# Distributed under the Boost Software License, Version 1.0. +# (See accompanying file LICENSE.txt or copy at +# https://www.bfgroup.xyz/b2/LICENSE.txt) + +from __future__ import print_function + +import BoostBuild + +import os +import os.path +import sys + +xml = "--xml" in sys.argv +toolset = BoostBuild.get_toolset() + + +# Clear environment for testing. +# +for s in ("BOOST_ROOT", "BOOST_BUILD_PATH", "JAM_TOOLSET", "BCCROOT", + "MSVCDir", "MSVC", "MSVCNT", "MINGW", "watcom"): + try: + del os.environ[s] + except: + pass + +BoostBuild.set_defer_annotations(1) + + +def run_tests(critical_tests, other_tests): + """ + Runs first the critical_tests and then the other_tests. + + Writes the name of the first failed test to test_results.txt. Critical + tests are run in the specified order, other tests are run starting with the + one that failed first on the last test run. + + """ + last_failed = last_failed_test() + other_tests = reorder_tests(other_tests, last_failed) + all_tests = critical_tests + other_tests + + invocation_dir = os.getcwd() + max_test_name_len = 10 + for x in all_tests: + if len(x) > max_test_name_len: + max_test_name_len = len(x) + + pass_count = 0 + failures_count = 0 + + for test in all_tests: + if not xml: + s = "%%-%ds :" % max_test_name_len % test + print(s, end='') + + passed = 0 + try: + __import__(test) + passed = 1 + except KeyboardInterrupt: + """This allows us to abort the testing manually using Ctrl-C.""" + raise + except SystemExit as e: + """This is the regular way our test scripts are supposed to report + test failures.""" + if e.code is None or e.code == 0: + passed = 1 + except: + exc_type, exc_value, exc_tb = sys.exc_info() + try: + BoostBuild.annotation("failure - unhandled exception", "%s - " + "%s" % (exc_type.__name__, exc_value)) + BoostBuild.annotate_stack_trace(exc_tb) + finally: + # Explicitly clear a hard-to-garbage-collect traceback + # related reference cycle as per documented sys.exc_info() + # usage suggestion. + del exc_tb + + if passed: + pass_count += 1 + else: + failures_count += 1 + if failures_count == 1: + f = open(os.path.join(invocation_dir, "test_results.txt"), "w") + try: + f.write(test) + finally: + f.close() + + # Restore the current directory, which might have been changed by the + # test. + os.chdir(invocation_dir) + + if not xml: + if passed: + print("PASSED") + else: + print("FAILED") + BoostBuild.flush_annotations() + else: + rs = "succeed" + if not passed: + rs = "fail" + print(''' +<test-log library="build" test-name="%s" test-type="run" toolset="%s" test-program="%s" target-directory="%s"> +<run result="%s">''' % (test, toolset, "tools/build/v2/test/" + test + ".py", + "boost/bin.v2/boost.build.tests/" + toolset + "/" + test, rs)) + if not passed: + BoostBuild.flush_annotations(1) + print(''' +</run> +</test-log> +''') + sys.stdout.flush() # Makes testing under emacs more entertaining. + BoostBuild.clear_annotations() + + # Erase the file on success. + if failures_count == 0: + open("test_results.txt", "w").close() + + if not xml: + print(''' + === Test summary === + PASS: %d + FAIL: %d + ''' % (pass_count, failures_count)) + + # exit with failure with failures + if failures_count > 0: + sys.exit(1) + +def last_failed_test(): + "Returns the name of the last failed test or None." + try: + f = open("test_results.txt") + try: + return f.read().strip() + finally: + f.close() + except Exception: + return None + + +def reorder_tests(tests, first_test): + try: + n = tests.index(first_test) + return [first_test] + tests[:n] + tests[n + 1:] + except ValueError: + return tests + + +critical_tests = ["unit_tests", "module_actions", "startup_v2", "core_d12", + "core_typecheck", "core_delete_module", "core_language", "core_arguments", + "core_varnames", "core_import_module"] + +# We want to collect debug information about the test site before running any +# of the tests, but only when not running the tests interactively. Then the +# user can easily run this always-failing test directly to see what it would +# have returned and there is no need to have it spoil a possible 'all tests +# passed' result. +if xml: + critical_tests.insert(0, "collect_debug_info") + +tests = ["abs_workdir", + "absolute_sources", + "alias", + "alternatives", + "always", + "bad_dirname", + "build_dir", + "build_file", + "build_hooks", + "build_no", + "builtin_echo", + "builtin_exit", + "builtin_glob", + "builtin_readlink", + "builtin_split_by_characters", + "bzip2", + "c_file", + "chain", + "clean", + "cli_property_expansion", + "command_line_properties", + "composite", + "conditionals", + "conditionals2", + "conditionals3", + "conditionals4", + "conditionals_multiple", + "configuration", + "configure", + "copy_time", + "core_action_output", + "core_action_status", + "core_actions_quietly", + "core_at_file", + "core_bindrule", + "core_dependencies", + "core_syntax_error_exit_status", + "core_fail_expected", + "core_jamshell", + "core_modifiers", + "core_multifile_actions", + "core_nt_cmd_line", + "core_option_d2", + "core_option_l", + "core_option_n", + "core_parallel_actions", + "core_parallel_multifile_actions_1", + "core_parallel_multifile_actions_2", + "core_scanner", + "core_source_line_tracking", + "core_update_now", + "core_variables_in_actions", + "custom_generator", + "debugger", +# Newly broken? +# "debugger-mi", + "default_build", + "default_features", +# This test is known to be broken itself. +# "default_toolset", + "dependency_property", + "dependency_test", + "disambiguation", + "dll_path", + "double_loading", + "duplicate", + "example_libraries", + "example_make", + "exit_status", + "expansion", + "explicit", + "feature_cxxflags", + "feature_implicit_dependency", + "feature_relevant", + "feature_suppress_import_lib", + "file_types", + "flags", + "generator_selection", + "generators_test", + "implicit_dependency", + "indirect_conditional", + "inherit_toolset", + "inherited_dependency", + "inline", + "install_build_no", + "libjpeg", + "liblzma", + "libpng", + "libtiff", + "libzstd", + "lib_source_property", + "lib_zlib", + "library_chain", + "library_property", + "link", + "load_order", + "loop", + "make_rule", + "message", + "ndebug", + "no_type", + "notfile", + "ordered_include", +# FIXME: Disabled due to bug in B2 +# "ordered_properties", + "out_of_tree", + "package", + "param", + "path_features", + "prebuilt", + "preprocessor", + "print", + "project_dependencies", + "project_glob", + "project_id", + "project_root_constants", + "project_root_rule", + "project_test3", + "project_test4", + "property_expansion", +# FIXME: Disabled due lack of qt5 detection +# "qt5", + "rebuilds", + "relative_sources", + "remove_requirement", + "rescan_header", + "resolution", + "rootless", + "scanner_causing_rebuilds", + "searched_lib", + "skipping", + "sort_rule", + "source_locations", + "source_order", + "space_in_path", + "stage", + "standalone", + "static_and_shared_library", + "suffix", + "tag", + "test_rc", + "test1", + "test2", + "testing", + "timedata", + "toolset_clang_darwin", + "toolset_clang_linux", + "toolset_clang_vxworks", + "toolset_darwin", + "toolset_defaults", + "toolset_gcc", + "toolset_intel_darwin", + "toolset_msvc", + "toolset_requirements", + "transitive_skip", + "unit_test", + "unused", + "use_requirements", + "using", + "wrapper", + "wrong_project", + ] + +if os.name == "posix": + tests.append("symlink") + # On Windows, library order is not important, so skip this test. Besides, + # it fails ;-). Further, the test relies on the fact that on Linux, one can + # build a shared library with unresolved symbols. This is not true on + # Windows, even with cygwin gcc. + +# Disable this test until we figure how to address failures due to --as-needed being default now. +# if "CYGWIN" not in os.uname()[0]: +# tests.append("library_order") + +if toolset.startswith("gcc") and os.name != "nt": + # On Windows it's allowed to have a static runtime with gcc. But this test + # assumes otherwise. Hence enable it only when not on Windows. + tests.append("gcc_runtime") + +if toolset.startswith("clang") or toolset.startswith("gcc") or toolset.startswith("msvc"): + tests.append("pch") + if sys.platform != "darwin": # clang-darwin does not yet support + tests.append("feature_force_include") + +# Clang includes Objective-C driver everywhere, but GCC usually in a separate gobj package +if toolset.startswith("clang") or "darwin" in toolset: + tests.append("lang_objc") + +# Disable on OSX as it doesn't seem to work for unknown reasons. +if sys.platform != 'darwin': + tests.append("builtin_glob_archive") + +if "--extras" in sys.argv: + tests.append("boostbook") + tests.append("qt4") + tests.append("qt5") + tests.append("example_qt4") + # Requires ./whatever.py to work, so is not guaranteed to work everywhere. + tests.append("example_customization") + # Requires gettext tools. + tests.append("example_gettext") +elif not xml: + print("Note: skipping extra tests") + +run_tests(critical_tests, tests) |