it's deprecated. Apple marked bootstrap_register as deprecated on 10.5 but
it's actually still needed on that OS release. There isn't a way to get the
functionality Breakpad needs from it without calling it until 10.6.
Review URL: https://breakpad.appspot.com/350001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@921 4c0a9323-5329-0410-9bdc-e9ce6186880e
This CL adds a minidump_generator that can write a minidump from a NSException
on iOS on an ARM cpu.
This CL also install an uncaught exception handler on iOS, and use the
previous generator to write minidumps for any uncaught exception.
Review URL: https://breakpad.appspot.com/347001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@916 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch allows Breakpad's DWARF reader to at least read or skip
attributes using the new forms defined in version 4 of the DWARF
specification, instead of crashing.
Attributes encoded using DW_FORM_flag_present, DW_FORM_sec_offset, and
DW_FORM_exprloc should work fine now. However, compilation units using
DW_FORM_ref_sig8 to refer to types in .debug_types will need further work
to support. (GCC 4.6.2 does not emit .debug_types sections.)
Specifically:
- dwarf2reader::DwarfForm gets new values.
- dwarf2reader::Dwarf2Handler and dwarf2reader::DIEHandler get new handler
methods, named ProcessAttributeSignature, for DW_FORM_ref_sig8 attributes.
- dwarf2reader::CompilationUnit reads DW_FORM_ref_sig8 attributes, and
passes them to ProcessAttributeSignature. It also gets support for
DW_FORM_sec_offset, DW_FORM_exprloc, and DW_FORM_flag_present, using the
existing appropriate ProcessAttribute* methods.
- dwarf2reader::DIEDispatcher passes through ProcessAttributeSignature
attributes to its DIEHandler.
- Unit tests are updated.
a=jimb, r=ted.mielczarek
Review URL: http://breakpad.appspot.com/343003/
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@912 4c0a9323-5329-0410-9bdc-e9ce6186880e
This is really incomplete --- it's just what's needed to get started
testing support for the DWARF 4 attribute forms.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@910 4c0a9323-5329-0410-9bdc-e9ce6186880e
When a variable is used to set (and lookup) MappingInfo's "start_addr"
field, it needs to match types -- which is "uintptr_t". When Chrome OS
updated the 'make' that's used for building, the 32-bit "char *" had
sign-extended when cast up to a u_int64_t -- maybe because pointers were
unsigned before and then changed to be signed -- and that caused the address
lookup to fail.
BUG=chromium-os:25355
TEST=Ran Breakpad unittests
A=mkrebs@chromium.org
Review URL: http://breakpad.appspot.com/345001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@908 4c0a9323-5329-0410-9bdc-e9ce6186880e
uploader.mm:549:5: warning: instance method '-uploadData:name:url:' not found (return type defaults to 'id')
[self uploadData:logFileData_ name:@"log" url:url];
=> it looks like this method does in fact not exist, the last parameter
needs to be removed.
breakpad_nlist_64.cc:193:59: warning: '&&' within '||' [-Wlogical-op-parentheses]
=> Just add parentheses, no functionality change.
Patch by Nico Weber <thakis@chromium.org>
BUG=none
TEST=breakpad stil works.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@907 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
Specifically, this patch makes the following changes:
1. Turn the LinuxDumper class into a base class and move ptrace related
code into a new derived class, LinuxPtraceDumper.
2. Add a LinuxCoreDumper class, which is derived from LinuxDumper, to
extract information from a crashed process via a core dump file instead
of ptrace.
3. Add a WriteMinidumpFromCore function to
src/client/linux/minidump_writer/minidump_writer.h,
which uses LinuxCoreDumper to extract information from a core dump file.
4. Add a core2md utility, which simply wraps WriteMinidumpFromCore, for
converting a core dump to a minidump.
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
4. Run Chromium OS tests to test core2md.
Review URL: http://breakpad.appspot.com/343001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@905 4c0a9323-5329-0410-9bdc-e9ce6186880e
CrashGenerator::CreateChildCrash() may have some flakiness. This patch
changes ElfCoreDumpTest to temporarily skip the ValidCoreFile test if
no core dump is generated by CrashGenerator::CreateChildCrash(), but
print out the error message to help debug the flakiness.
BUG=chromium-os:24982
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.6.
2. All unit tests pass.
Review URL: http://breakpad.appspot.com/342001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@904 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
Specifically, this patch makes the following changes:
1. Add two convenient methods, back() and empty(), to the wasteful_vector
class.
2. Refactor the LinuxDumper class such that it can later be splitted into
a base class and two derived classes, one uses the current ptrace
implementation and one uses a core file.
3. Refactor the MinidumpWriter class such that it can later use different
derived implementations of LinuxDumper.
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
Review URL: http://breakpad.appspot.com/340001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@902 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
Specifically, this patch makes the following changes:
1. Add an ElfCoreDump class for processing Linux core dump files, which will
later be used to implement the core dump to minidump conversion.
2. Add a CrashGenerator class for generating a crash with a core dump file
for testing the functionalities of ElfCoreDump.
3. Move some utility functions for reading/writing files to file_utils.h.
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
Review URL: http://breakpad.appspot.com/337001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@900 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
The MemoryRange class was added in r895 (http://breakpad.appspot.com/332001),
which is largely based on MMappedRange but generalized to be used in other
code. However, MemoryRange does not support minidump data structures. This
patch adds a MinidumpMemoryRange class that extends MemoryRange to handle
minidump data structures, which can then replace MMappedRange.
As with MemoryRange, MinidumpMemoryRange is unit tested.
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
4. Run minidump-2-core to covnert a minidump file to a core file.
Review URL: http://breakpad.appspot.com/335001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@898 4c0a9323-5329-0410-9bdc-e9ce6186880e
Manufacturing an exception record improves the crash reporting, since then
the crashes get bucketed by the call stack leading to the dump, instead of
all falling into a misc bucket that nobody ever looks at.
Currently these are occuring through e.g. dumps from the base watchdog.
Link against RtlCaptureContext, as the function has been documented as
available
from Windows XP
[http://msdn.microsoft.com/en-us/library/windows/desktop/ms680591(v=vs.85).aspx].
Patch by Siggi Asgeirsson <siggi@chromium.org>
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@897 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
Specifically, this patch makes the following changes:
1. Add a SafeReadLink function that wraps sys_readlink() to resolve a
symbolic link but guarantees the result is NULL-terminated on success.
2. Refactor other source code to use SafeReadLink instead of readlink()
or sys_readlink().
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
4. Run minidump-2-core to covnert a minidump file to a core file.
Review URL: http://breakpad.appspot.com/334001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@896 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch is part of a bigger patch that helps merging the breakpad code
with the modified version in Chromium OS.
Specifically, this patch makes the following changes:
1. Add a MemoryRange class for encapsulating and checking read access
to a contiguous range of memory.
2. Add a MemoryMappedFile class for mapping a file into memory for
read-only access.
3. Refactor other source code to use MemoryMappedFile.
BUG=455
TEST=Tested the following:
1. Build on 32-bit and 64-bit Linux with gcc 4.4.3 and gcc 4.6.
2. Build on Mac OS X 10.6.8 with gcc 4.2 and clang 3.0 (with latest gmock).
3. All unit tests pass.
4. Run minidump-2-core to covnert a minidump file to a core file.
Review URL: http://breakpad.appspot.com/332001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@895 4c0a9323-5329-0410-9bdc-e9ce6186880e
The assertion (context.context_flags & MD_CONTEXT_X86) in synth_minidump.cc
produces a different message on Mac OS X than other platforms. This patch
changes the ContextDeathTest.X86BadFlags test to handle both message patterns.
BUG=424
TEST=Verified ContextDeathTest.X86BadFlags on Windows, Linux and Mac OS X.
Review URL: http://breakpad.appspot.com/333001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@894 4c0a9323-5329-0410-9bdc-e9ce6186880e
"const" that causes an error with the newer gcc.
BUG=chromium-os:14973
TEST=Built CrOS's google-breakpad
Patch by Michael Krebs <mkrebs@chromium.org>
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@878 4c0a9323-5329-0410-9bdc-e9ce6186880e
To be noted:
1) All is done in process, as multi-process is not allowed on ios.
2) Dump are saved when a crash occures but are not automatically send to the
server.
3) Breakpad.h contains function to check if a dump must be uploaded, and to
upload a dump.
4) The code is copy pasting a log of Breakpad implementation for Mac OS. It
might be possible to do some refactoring.
Review URL: http://breakpad.appspot.com/309003
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@868 4c0a9323-5329-0410-9bdc-e9ce6186880e
forbidden by the style guide, and are bad practice in headers even under
style rules that tolerate this construct.
This fixes warnings such as:
In file included from dwarf2reader.cc:34:
dwarf2reader.h:53:17: warning: using namespace directive in global context in header [-Wheader-hygiene]
Review URL: http://breakpad.appspot.com/312002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@862 4c0a9323-5329-0410-9bdc-e9ce6186880e
This function already establishes that (signed) i must be positive by the time
it assigns it to (unsigned) reg. Because reg is unsigned, it is impossible for
it to be negative. Because i was already checked for positivity, this check
can be removed entirely. It was not a miswritten check that intended to check
i instead of reg, because i is already checked.
This addresses the following warning:
dwarf_cfi_to_module.cc:135:9: warning: comparison of 0 <= unsigned expression is always true [-Wtautological-compare]
Committed r859
Review URL: http://breakpad.appspot.com/310003
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@861 4c0a9323-5329-0410-9bdc-e9ce6186880e
symupload printed -[NSData length], an NSUInteger, using %lu. %lu is proper
to print a "long" as unsigned, but NSUInteger is a typedef for "unsigned int"
when building for 32-bit. This would not have caused any problems, because in
the 32-bit model, both int and long are 32 bits wide. In the 64-bit model,
long is 64 bits wide, but NSUInteger is defiend as "unsigned long", so there
wouldn't have even been a warning in that case.
This addresses the following warning:
symupload.m:137:30:{137:28-137:31}{137:46-137:59}: warning: conversion specifies type 'unsigned long' but the argument has type 'NSUInteger' (aka 'unsigned int') [-Wformat]
Review URL: http://breakpad.appspot.com/313002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@860 4c0a9323-5329-0410-9bdc-e9ce6186880e
The actual defintion comes from common/mac/GTMLogger.h, #imported above.
This removes a compilation warning that showed up in r853:
uploader.m:52:9: warning: 'GTMLoggerDebug' macro redefined
GTMLogger.h:339:9: note: previous definition is here
Review URL: http://breakpad.appspot.com/310002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@859 4c0a9323-5329-0410-9bdc-e9ce6186880e
<Breakpad/BreakpadDefines.h>.
During the build of Breakpad itself (especially in Chromium), there is no
Breakpad.framework/Headers in which to locate BreakpadDefines.h, so the
framework-style #include of <Breakpad/BreakpadDefines.h> is not able to find
anything to #include. Using the "BreakpadDefines.h" form should always locate
this file next to Breakpad.h whether it's in the framework or in the source
tree.
This fixes a Chromium build regression caused by Breakpad r856.
Review URL: http://breakpad.appspot.com/313001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@857 4c0a9323-5329-0410-9bdc-e9ce6186880e
This will be needed for iOS implementation, where the Inspector won't be
used, but where a config file will still be needed, because the uploads
won't happen just after the crash, but on a next run.
Review URL: http://breakpad.appspot.com/309001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@852 4c0a9323-5329-0410-9bdc-e9ce6186880e
Regression from Breakpad r842 (Chromium r103778) - browser crash reports were
uploaded, but renderer crash reports were not. Messages such as these may have
been logged:
com.apple.launchd.peruser.x[y] could not lookup DNS configuration info
service: (ipc/send) invalid destination port
com.apple.launchd.peruser.x[y] Breakpad Reporter: Send Error: Error
Domain=NSURLErrorDomain Code=-1009 UserInfo=z "This computer’s Internet
connection appears to be offline." Underlying Error=(Error
Domain=kCFErrorDomainCFNetwork Code=-1009 UserInfo=w "This computer’s Internet
connection appears to be offline.")
When OnDemandServer establishes the bootstrap subset, it will now register the
parent bootstrap port in the subset namespace so that the Inspector can
recover this port and switch to it. The Sender, launched by the Inspector,
relies on the bootstrap port being set properly.
BUG=chromium:99252
TEST=All test cases from Chromium r103778 (bug chromium:28547) plus:
about:crash should generate a crash report which should be uploaded,
provided that throttling is not in effect. Remove or edit
~/Library/Preferences/com.Breakpad.crash_report_sender.plist to defeat
throttling. Also verify that about:inducebrowsercrashforrealz works.
Review URL: http://breakpad.appspot.com/307001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@843 4c0a9323-5329-0410-9bdc-e9ce6186880e
lifetime of the task to be monitored, the invoking task. This allows the
bootstrap server (in launchd) to automatically clean up the Mach server
registration when the task being monitored exits, avoiding leaks of
com.Breakpad.Inspector(pid) ports in "launchctl bslist".
BUG=chromium:28547
TEST=Handler should still crash catches, but inspector ports should no longer
show up in "launchctl bslist". They should show up under a subset port in
"launchctl bstree" instead. "launchctl bstree" must be invoked as root.
Review URL: http://breakpad.appspot.com/306001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@842 4c0a9323-5329-0410-9bdc-e9ce6186880e
was already the SDK being used for x86_64 Release mode. The 10.6 SDK is not
necessary.
Explicitly set the file encoding to UTF-16 on the sender app's lproj's
InfoPlist.strings and Localizable.strings files.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@840 4c0a9323-5329-0410-9bdc-e9ce6186880e
The value of MD_CONTEXT_CPU_MASK in use assumes that only the lower 6 bits are used for flags, and the upper 26 bits are for the CPU type. However, as of Windows 7 SP1, the 7th bit is being used as a flag (per http://msdn.microsoft.com/en-us/library/hh134238%28v=vs.85%29.aspx and the Windows SDK headers). Adjusting MD_CONTEXT_CPU_MASK works, but unfortunately that masks off the existing value of MD_CONTEXT_ARM. This patch also changes the value of MD_CONTEXT_ARM and adjusts the minidump context reading machinery to gracefully handle minidumps with the old value.
R=mark at http://breakpad.appspot.com/302001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@831 4c0a9323-5329-0410-9bdc-e9ce6186880e
effect.
BUG=none
TEST=Apple Crash Reporter logs from processes in which Breakpad handles the
crash should point the finger at the actual crash source, not the
Breakpad thread's attempt to write to unwritable memory.
Review URL: http://breakpad.appspot.com/301001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@828 4c0a9323-5329-0410-9bdc-e9ce6186880e
split up into multiple regions.
An older workaround relyied on known fixed stack locations and only filled in
the initial page of the stack if it was in a distinct region. The new approach
looks upwards for additional regions that appear to be part of the same stack.
With PIE on Lion, the stack no longer begins at a fixed address, so the older
workaround became ineffective.
BUG=247, chromium:94107
TEST=Stacks should run through to _main/start and then stop when examining
Chrome on Lion with PIE and "slid" stacks.
Review URL: http://breakpad.appspot.com/300001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@826 4c0a9323-5329-0410-9bdc-e9ce6186880e
This libcrypto dependency sucks. Linking against OpenSSL is sort of broken in
certain Mac OS X SDKs. libcrypto was only being used to provide an MD5
implementation. Breakpad already has its own MD5 implementation, so just use
that instead.
To be perfectly honest, on modern systems, nothing should be making MD5
hashes of modules anyway, because everything has an embedded LC_UUID.
The project file changes just remove libcrypto and add md5.c as needed.
A bonus (and untested) fix for on_demand_symbol_supplier.mm is included to
account for changes in r794.
Review URL: http://breakpad.appspot.com/296001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@819 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch adds synth_elf::{StringTable,SymbolTable,ELF} classes to
produce in-memory ELF files to properly test the Linux symbol dumping
code. It also uses those classes to add some basic tests for
the WriteSymbolFile function.
R=jimb at http://breakpad.appspot.com/277001/show
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@794 4c0a9323-5329-0410-9bdc-e9ce6186880e
Fix an assertion where a zero-length buffer was being passed to
UntypedMDRVA::Copy(). This occurred when WriteFile() was given a file whose
size was a multiple of the temporary buffer size. In this issue's case, the
procfs file "environ" happened to be 2032 bytes, while the temporary buffer
was 1016 bytes.
Patch by Michael Krebs <mkrebs@chromium.org>
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@792 4c0a9323-5329-0410-9bdc-e9ce6186880e
In r779, at the last moment, I added a default call count expectation for
the UnnamedFunction warning to the CUFixtureBase constructor, but didn't
re-run the tests. This patch adjusts all affected tests.
a=jimblandy, r=ted.mielczarek
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@782 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch makes sure dump_syms behaves properly when presented with
malformed DWARF data that provides no name for a function. We print a
warning message to stderr, and subsitute "<name omitted>" for the empty
string, so that the "FUNC" record written to the symbol file for the
function is still well-formed. (We may have line number data covering the
function, so it would be a shame to omit the function altogether.)
Unit tests included.
a=jimblandy, r=ted.mielczarek
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@779 4c0a9323-5329-0410-9bdc-e9ce6186880e
is no reason not to keep it locally. Implemented a basic disassembler which can be used
to scan bytecode for interesting conditions. This should be pretty easy to add to for
things other than exploitability if there is a desire. This also adds several tests to
the windows exploitability ranking code to take advantage of the disassembler for x86
code.
BUG=None
TEST=DisassemblerX86Test.*
Review URL: http://breakpad.appspot.com/203001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@705 4c0a9323-5329-0410-9bdc-e9ce6186880e
Backed out r684 (added glog include dir to client gyp files). It was obviated by r685, which removed the dependency on glog from the client projects.
BUG=None
TEST="gclient runhooks --force"; build crash_generation_app; launch crash_generation_app.
r=hansl at http://breakpad.appspot.com/191001/show
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@696 4c0a9323-5329-0410-9bdc-e9ce6186880e
After the final DIE in a compilation unit, there may be any number of
zero bytes present. This is meant to allow producers to align
compilation unit starting points when necessary.
This patch changes the dwarf2reader::CompilationUnit class to skip
those zero bytes, rather than interpreting them as 'end of children'
markers for DIEs that do not exist. Without this change, the padding
bytes will cause the reader to attempt to pop an offset from an empty
stack, and call EndDIE with a garbage offset.
a=jimblandy, r=mmentovai
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@667 4c0a9323-5329-0410-9bdc-e9ce6186880e
Perhaps there once was some reason one needed the DIE offset stack to
have an unusual lifetime, but there is none now.
a=jimblandy, r=mmentovai
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@666 4c0a9323-5329-0410-9bdc-e9ce6186880e
tgkill() is not necessarily possible, as a sandbox might block this call.
This changelist tries different approaches depending on whether we received
a synchronous or an asynchronous signal. This fixes unittest failures and
also runs correctly in sandbox'd environments.
TEST=ran unittest, and opened about:crash in sandbox'd Chrome
BUG=395
A=markus@chromium.org
Original review: http://breakpad.appspot.com/159001
Review URL: http://breakpad.appspot.com/146002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@656 4c0a9323-5329-0410-9bdc-e9ce6186880e
what architecture name is shown in a symbol file's MODULE line, but the Mac
crash_report tool's on_demand_symbol_supplier does. The new Mac dumper
inadvertently used i386. Correct that to make it x86. Temporarily make the
on_demand_symbol_supplier accept symbol files whose architecture is i386.
Also add x86_64 to the set of architectures that the on_demand_symbol_supplier
considers valid.
BUG=none
TEST=none
Review URL: http://breakpad.appspot.com/143001
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@638 4c0a9323-5329-0410-9bdc-e9ce6186880e
Inspector::ReadMessages as was done before r627. The "hello" message contains
the parameter count and is referenced while the message reader loops through
parameter messages. Prior to r627, both messages were named |message|, which
was confusing, probably caused a compiler warning, and apparently provided the
motivation to share them. This caused the crash inspector to fail to properly
collect the parameters. The common failure mode (although others are possible)
was for the inspector to attempt tor read more parameter messages than were
available, resulting in an IPC timeout and inspector death. No crash report
would be written, and the application expecting its crash to be inspected
would time out waiting for a response from the inspector and then _exit. This
is effectively a failure to properly handle crashes.
The inner message is reintroduced, and named parameter_message for
disambiguation.
BUG=chromium:49821
TEST=Crashes catchable by the Mac Breakpad framework
Review URL: http://breakpad.appspot.com/123002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@628 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch avoids allocating many copies of identical strings appearing in
debugging information. Without this patch, running dump_syms on Mozilla's
libxul.so (with 173MiB of debugging information) has a peak resident set of
around 450MiB. With this patch, the peak is around 365MiB.
a=jimblandy, r=mark
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@626 4c0a9323-5329-0410-9bdc-e9ce6186880e
At present, the Linux symbol dumper maps the ELF file into memory to
examine the debugging information it contains, but then also calls
google_breakpad::FileID::ElfFileIdentifier, which maps the ELF file into
memory again. Some of our object files are large; Mozilla's libxul.so is
1.1GiB. Trying to map such files twice can interfere with tools like
valgrind that map themselves into high addresses (in an attempt to stay out
of the way of ordinary programs).
The FileID class has another method, ElfFileIdentifierFromMappedFile, that
operates on an already-loaded image of the file; use that instead.
a=jimblandy, r=thestig
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@625 4c0a9323-5329-0410-9bdc-e9ce6186880e
