summaryrefslogtreecommitdiffstats
path: root/testing/mozbase/mozleak/mozleak/lsan.py
blob: f6555eff2d8074e7bff96649687d4bfb62678f44 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
# 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 re


class LSANLeaks(object):

    """
    Parses the log when running an LSAN build, looking for interesting stack frames
    in allocation stacks
    """

    def __init__(
        self,
        logger,
        scope=None,
        allowed=None,
        maxNumRecordedFrames=None,
        allowAll=False,
    ):
        self.logger = logger
        self.inReport = False
        self.fatalError = False
        self.symbolizerError = False
        self.foundFrames = set()
        self.recordMoreFrames = None
        self.currStack = None
        self.maxNumRecordedFrames = maxNumRecordedFrames if maxNumRecordedFrames else 4
        self.summaryData = None
        self.scope = scope
        self.allowedMatch = None
        self.allowAll = allowAll
        self.sawError = False

        # Don't various allocation-related stack frames, as they do not help much to
        # distinguish different leaks.
        unescapedSkipList = [
            "malloc",
            "js_malloc",
            "malloc_",
            "__interceptor_malloc",
            "moz_xmalloc",
            "calloc",
            "js_calloc",
            "calloc_",
            "__interceptor_calloc",
            "moz_xcalloc",
            "realloc",
            "js_realloc",
            "realloc_",
            "__interceptor_realloc",
            "moz_xrealloc",
            "new",
            "js::MallocProvider",
        ]
        self.skipListRegExp = re.compile(
            "^" + "|".join([re.escape(f) for f in unescapedSkipList]) + "$"
        )

        self.startRegExp = re.compile(
            "==\d+==ERROR: LeakSanitizer: detected memory leaks"
        )
        self.fatalErrorRegExp = re.compile(
            "==\d+==LeakSanitizer has encountered a fatal error."
        )
        self.symbolizerOomRegExp = re.compile(
            "LLVMSymbolizer: error reading file: Cannot allocate memory"
        )
        self.stackFrameRegExp = re.compile("    #\d+ 0x[0-9a-f]+ in ([^(</]+)")
        self.sysLibStackFrameRegExp = re.compile(
            "    #\d+ 0x[0-9a-f]+ \(([^+]+)\+0x[0-9a-f]+\)"
        )
        self.summaryRegexp = re.compile(
            "SUMMARY: AddressSanitizer: (\d+) byte\(s\) leaked in (\d+) allocation\(s\)."
        )
        self.rustRegexp = re.compile("::h[a-f0-9]{16}$")
        self.setAllowed(allowed)

    def setAllowed(self, allowedLines):
        if not allowedLines or self.allowAll:
            self.allowedRegexp = None
        else:
            self.allowedRegexp = re.compile(
                "^" + "|".join([re.escape(f) for f in allowedLines])
            )

    def log(self, line):
        if re.match(self.startRegExp, line):
            self.inReport = True
            # Downgrade this from an ERROR
            self.sawError = True
            return "LeakSanitizer: detected memory leaks"

        if re.match(self.fatalErrorRegExp, line):
            self.fatalError = True
            return line

        if re.match(self.symbolizerOomRegExp, line):
            self.symbolizerError = True
            return line

        if not self.inReport:
            return line

        if line.startswith("Direct leak") or line.startswith("Indirect leak"):
            self._finishStack()
            self.recordMoreFrames = True
            self.currStack = []
            return line

        summaryData = self.summaryRegexp.match(line)
        if summaryData:
            assert self.summaryData is None
            self._finishStack()
            self.inReport = False
            self.summaryData = (int(item) for item in summaryData.groups())
            # We don't return the line here because we want to control whether the
            # leak is seen as an expected failure later
            return

        if not self.recordMoreFrames:
            return line

        stackFrame = re.match(self.stackFrameRegExp, line)
        if stackFrame:
            # Split the frame to remove any return types.
            frame = stackFrame.group(1).split()[-1]
            if not re.match(self.skipListRegExp, frame):
                self._recordFrame(frame)
            return line

        sysLibStackFrame = re.match(self.sysLibStackFrameRegExp, line)
        if sysLibStackFrame:
            # System library stack frames will never match the skip list,
            # so don't bother checking if they do.
            self._recordFrame(sysLibStackFrame.group(1))

        # If we don't match either of these, just ignore the frame.
        # We'll end up with "unknown stack" if everything is ignored.
        return line

    def process(self):
        failures = 0

        if self.allowAll:
            self.logger.info("LeakSanitizer | Leak checks disabled")
            return

        if self.summaryData:
            allowed = all(allowed for _, allowed in self.foundFrames)
            self.logger.lsan_summary(*self.summaryData, allowed=allowed)
            self.summaryData = None

        if self.fatalError:
            self.logger.error(
                "LeakSanitizer | LeakSanitizer has encountered a fatal error."
            )
            failures += 1

        if self.symbolizerError:
            self.logger.error(
                "LeakSanitizer | LLVMSymbolizer was unable to allocate memory.\n"
                "This will cause leaks that "
                "should be ignored to instead be reported as an error"
            )
            failures += 1

        if self.foundFrames:
            self.logger.info(
                "LeakSanitizer | To show the "
                "addresses of leaked objects add report_objects=1 to LSAN_OPTIONS\n"
                "This can be done in testing/mozbase/mozrunner/mozrunner/utils.py"
            )
            self.logger.info("Allowed depth was %d" % self.maxNumRecordedFrames)

            for frames, allowed in self.foundFrames:
                self.logger.lsan_leak(frames, scope=self.scope, allowed_match=allowed)
                if not allowed:
                    failures += 1

        if self.sawError and not (
            self.summaryData
            or self.foundFrames
            or self.fatalError
            or self.symbolizerError
        ):
            self.logger.error(
                "LeakSanitizer | Memory leaks detected but no leak report generated"
            )

        self.sawError = False

        return failures

    def _finishStack(self):
        if self.recordMoreFrames and len(self.currStack) == 0:
            self.currStack = {"unknown stack"}
        if self.currStack:
            self.foundFrames.add((tuple(self.currStack), self.allowedMatch))
            self.currStack = None
            self.allowedMatch = None
        self.recordMoreFrames = False
        self.numRecordedFrames = 0

    def _recordFrame(self, frame):
        if self.allowedMatch is None and self.allowedRegexp is not None:
            self.allowedMatch = frame if self.allowedRegexp.match(frame) else None
        frame = self._cleanFrame(frame)
        self.currStack.append(frame)
        self.numRecordedFrames += 1
        if self.numRecordedFrames >= self.maxNumRecordedFrames:
            self.recordMoreFrames = False

    def _cleanFrame(self, frame):
        # Rust frames aren't properly demangled and in particular can contain
        # some trailing junk of the form ::h[a-f0-9]{16} that changes with
        # compiler versions; see bug 1507350.
        return self.rustRegexp.sub("", frame)