- Resets `selected_object_file_` when a new file is read. This was a dangling pointer previously.
- When `-a` is provided, ensures that both parts of a split module use
the given architecture.
Bug: None
Change-Id: I581d41b0eee4ec2b0d598fb80b9065e7ebde0e0d
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3788222
Reviewed-by: Mark Mentovai <mark@chromium.org>
DW_CFA_AARCH64_negate_ra_state uses the same encoding as
DW_CFA_GNU_window_save. It is meant to indicate that x30/LR
has been signed, but breakpad does not require this information.
Bug: b/239086293
Change-Id: I5a17bd5e0673a3ff80a8c6e347013d66054314e8
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3781136
Reviewed-by: Mark Mentovai <mark@chromium.org>
Previously these all resulted in EXCEPTION_STACK_BUFFER_OVERRUN
but this hides various specific fast fail crash types, which
are now provided based on the exception's subcode.
Tests: added to minidump_process_unittest.cc
Bug: 865632
Change-Id: Ic6693de247da55cf6d132d108c6e20c635f366b1
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3771437
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Explicitly call out where overflows are expected, and add appropriate
checking for them.
BUG=b:235999011
TEST=Unittests on CrOS and Linux
Change-Id: I999a6996183c2f4afc16a1c0188dee3bd64d7f09
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3759630
Reviewed-by: Mike Frysinger <vapier@chromium.org>
On ARM, this write fails to build:
comparison of integers of different signs: 'ssize_t' (aka 'int') and
'const unsigned int' [-Werror,-Wsign-compare]
Since we check that it's <= 15 above, we can simply cast it without
issue.
Bug: b:235999011
Change-Id: Id75fc0df74e88b347df615df06567e96c6b59a1d
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3758800
Reviewed-by: Mike Frysinger <vapier@chromium.org>
This test's constructor fails to fully initialize this raw context. As a
result, we have at least one use of uninitialized memory in CFI.At4004.
This causes said test to fail under ASAN.
Bug: b:235999011
Change-Id: I3279da8ac3414eb8c90f3949a1db47a03e750a94
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3756749
Reviewed-by: Mike Frysinger <vapier@chromium.org>
`AddFunction` only takes ownership of the pointer passed to it if it
returns true. Since it returns false when adding `function2`, we need to
free it.
Bug: b:235999011
Change-Id: I11984103c2c153ff0daf2c9690f9c88d04a2131b
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3756732
Reviewed-by: Mike Frysinger <vapier@chromium.org>
Everything in `functions_` is owned by the current `StabsToModule`. If
we fail to add something from `functions_`, we need to be sure to
dispose of it properly, since `module_` will not take ownership.
Bug: b:235999011
Change-Id: I3b965709ea2016a065b50588f4132d14a1de7725
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3756733
Reviewed-by: Mike Frysinger <vapier@chromium.org>
These `GoogleCrashdumpUploader` instances need to be cleaned up; place
them on the stack.
Doing this unmasks another bug in this code: the `MockLibcurlWrapper`
instance we're passing into these `GoogleCrashdumpUploader`s becomes
owned by the `GoogleCrashdumpUploader` in question. Putting them on the
stack makes `free()` unhappy when the `GoogleCrashdumpUploader` they're
given to gets destructed.
Bug: b:235999011
Change-Id: I5d0424a1c09d32ea34a8fa6f5e52d3695ee6e857
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3756172
Reviewed-by: Mike Frysinger <vapier@chromium.org>
exploitability_linux assumed a 15 byte buffer to always be passed in as
`raw_bytes` for `DisassembleBytes`. This test was passing in a 6 byte
buffer. Make `DisassembleBytes` accept a length.
Bug: b:235999011
Change-Id: I696c66357faa1c7d762c64009864123897f03488
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3756170
Reviewed-by: Mike Frysinger <vapier@chromium.org>
On CrOS, a breakpad update is breaking:
```
google-breakpad-2022.07.03.054510-r210: no member named 'sort' in
namespace 'std'
google-breakpad-2022.07.03.054510-r210: std::sort(abbrevs_->begin(),
abbrevs_->end(),
google-breakpad-2022.07.03.054510-r210: ~~~~~^
```
Looks like we're missing the include for `sort`.
Bug: 235999011
Change-Id: I917389c12b370357fd1fc7cb08af0b9d7f315c84
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3741510
Reviewed-by: Mike Frysinger <vapier@chromium.org>
Currently, `./configure` supports
`--with-rustc-demangle=${path_to_crate_root}` as the only mechanism to
enable rustc-demangle support. This CL adds support for cases where
keeping or synthesizing a full `cargo` build tree is hacky in comparison
to simply installing `rustc_demangle.h` in e.g., /usr/include, and
putting `librustc_demangle.a` in e.g., `/usr/lib64`.
Bug: b:235999011
Change-Id: Id5fe2a24c4b6e33c4df0e10c86ba99c7cf890ab2
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3733672
Reviewed-by: Mike Frysinger <vapier@chromium.org>
There are cases where the debug_abbrev entities are not sequential,
for example, in Xamarin system dlls. This change gracefully handles
such a case.
Change-Id: Ib270393d3cf9fd18efd99d15d0fba4f96748188a
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3738879
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Leverage the existing default logic & argument parsing that turn the
values yes/no into true/false to simplify the code so we don't create
duplicate variables. This kills a lot of boilerplate.
Change-Id: Ib7c8e00f7b23e67ed05f3b35e523c235aed41129
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3734169
Reviewed-by: George Burgess <gbiv@chromium.org>
Previous updates to these were partially done. Rerun with the
right versions of autoconf-2.69 & automake-1.16.5.
Change-Id: Ifd6c8405b0b50c5d3cf4ea536a7db5762d62644e
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3734167
Reviewed-by: George Burgess <gbiv@chromium.org>
We're using autoconf-2.69 (from 2012) now and not testing anything
older, so raise the min version to that.
Also bump to automake-1.13 (from 2012) to keep a bit inline.
Change-Id: I903144f214fef835364474e1607a26e846c99ff6
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3734168
Reviewed-by: George Burgess <gbiv@chromium.org>
pe_file.cc, used in minidump_writer.cc, was not included in LOCAL_SRC_FILES. This caused breakpad to fail in build time, as it didn't find the file.
Tested using ndk 21.4.7075529, building on an arm64 device.
Change-Id: I192539a52c1344eba9999c6780bcd8b83ea7e772
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3723683
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
This expectation started failing at
8b68c72a3f. Remove it.
This is the only expectation that exists in this test, but the test
may be useful for internal asserts performed in the code under test.
Bug: b:235999011
Change-Id: Iab5c073161ce66fdf362b7da31c19f471c7a79bf
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3727478
Reviewed-by: Mike Frysinger <vapier@chromium.org>
`Construct.FunctionsWithSameAddress` started failing at ff5892c5. It
looks like the cause of this is in the calls to
`generate_duplicate_function`:
```
generate_duplicate_function("_without_form");
generate_duplicate_function("_and_void");
```
`generate_duplicate_function` directly calls `new
Module::Function(...);`, which stores the `StringView` it's given.
`generate_duplicate_function` currently takes a `const
string &`; in the above statements, these strings get `free()`d at the
`;`.
Making the parameter a `StringView` means the `Module::Function` will
store pointers to the string literal, which lives for the whole program.
All calls to `generate_duplicate_function` are given literals.
Bug: b:235999011
Change-Id: Ied04c1307a2467b9816a83f0c4d84d47779ec610
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3726855
Reviewed-by: Mike Frysinger <vapier@chromium.org>
Some systems have constrained rlimits for core files (the CrOS chroot is
an example of this). Fail gracefully in this case, rather than breaking
the user's tests.
Bug: b:235999011
Change-Id: I5649b42d3e6fd9b4f9b11e1fd9d0d4a1083d300f
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3722724
Reviewed-by: Mark Mentovai <mark@chromium.org>
Reviewed-by: Mike Frysinger <vapier@chromium.org>
Evaluating CFI rules may fail due to e.g. the unavailability of the
memory or some register values. Failures in recovering registers other
than CFA or return address can be ignored because they are optional.
Bug: fuchsia:102034
Change-Id: Ia1d8bdb12766e32b5445b49d353fc71c63ab73e7
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3701260
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
This is a reland of commit 0808030bee
Original change's description:
> Support PE modules in core files when running core2md
>
> Core files generated from `wine` contain both ELF and PE modules. Module
> format can be guessed by checking the file contents. If the module
> corresponds to PE-file conditions (has specific fields set up as
> described in https://code.google.com/archive/p/corkami/wikis/PE.wiki)
> we'll create a MDCVInfoPDB70 record in the minidump for it, but if
> the file cannot be opened, is too short or is not a PE file, we'll
> fall back to ELF procedure.
>
> Added /src/client/linux/minidump_writer/pe_file.{cc,h} to
> src_client_linux_libbreakpad_client_a_SOURCES and
> src_client_linux_linux_client_unittest_shlib_SOURCES.
> Makefile.in and aclocal.m4 were generated by running 'aclocal && automake'.
>
> Test: build core2md and use it to convert a core file into dmp, validate
> that the generated dmp file can be opened. Ran './configure & make'.
>
> Change-Id: I225ffeea3f582deed40ecdfe7ab77f5754e90cbe
> Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3629189
> Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Change-Id: I09dd067a39a95f81f48656595e811c263561ebf2
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3695863
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
sprintf is marked as deprecated with Xcode 14.
Bug: 1331345
Change-Id: Ic301134ec0c5e7b9ee9d590ab1423491aad5ccf7
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3692036
Reviewed-by: Ivan Penkov <ivanpe@chromium.org>
Reviewed-by: Mike Frysinger <vapier@chromium.org>
This reverts commit 0808030bee.
Reason for revert: Breaks Android Compile
ld.lld: error: undefined symbol: google_breakpad::PEFile::TryGetDebugInfo(char const*, google_breakpad::_RSDS_DEBUG_FORMAT*)
>>> referenced by minidump_writer.cc
>>> client/minidump_writer.o:((anonymous namespace)::MinidumpWriter::FillRawModule(google_breakpad::MappingInfo const&, bool, unsigned int, MDRawModule*, unsigned char const*)) in archive obj/third_party/breakpad/libclient.a
https://ci.chromium.org/ui/p/chromium/builders/try/android-marshmallow-arm64-rel/1188618/overview
Original change's description:
> Support PE modules in core files when running core2md
>
> Core files generated from `wine` contain both ELF and PE modules. Module
> format can be guessed by checking the file contents. If the module
> corresponds to PE-file conditions (has specific fields set up as
> described in https://code.google.com/archive/p/corkami/wikis/PE.wiki)
> we'll create a MDCVInfoPDB70 record in the minidump for it, but if
> the file cannot be opened, is too short or is not a PE file, we'll
> fall back to ELF procedure.
>
> Added /src/client/linux/minidump_writer/pe_file.{cc,h} to
> src_client_linux_libbreakpad_client_a_SOURCES and
> src_client_linux_linux_client_unittest_shlib_SOURCES.
> Makefile.in and aclocal.m4 were generated by running 'aclocal && automake'.
>
> Test: build core2md and use it to convert a core file into dmp, validate
> that the generated dmp file can be opened. Ran './configure & make'.
>
> Change-Id: I225ffeea3f582deed40ecdfe7ab77f5754e90cbe
> Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3629189
> Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Change-Id: I7105ed615a338263f112243bd8dc9e86b906fcb1
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3695862
Reviewed-by: Ivan Penkov <ivanpe@chromium.org>
Core files generated from `wine` contain both ELF and PE modules. Module
format can be guessed by checking the file contents. If the module
corresponds to PE-file conditions (has specific fields set up as
described in https://code.google.com/archive/p/corkami/wikis/PE.wiki)
we'll create a MDCVInfoPDB70 record in the minidump for it, but if
the file cannot be opened, is too short or is not a PE file, we'll
fall back to ELF procedure.
Added /src/client/linux/minidump_writer/pe_file.{cc,h} to
src_client_linux_libbreakpad_client_a_SOURCES and
src_client_linux_linux_client_unittest_shlib_SOURCES.
Makefile.in and aclocal.m4 were generated by running 'aclocal && automake'.
Test: build core2md and use it to convert a core file into dmp, validate
that the generated dmp file can be opened. Ran './configure & make'.
Change-Id: I225ffeea3f582deed40ecdfe7ab77f5754e90cbe
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3629189
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
https://crrev.com/c/3327644 introduced the ability for dump_syms to
operate on in memory data, which has the consequence of not going
through the same input validation as the dump_syms cli tool. In certain
cases, it is possible that architecture info can't be reliably
determined, e.g. new architectures that breakpad is unware of. In that
case, dump_syms should avoid crashing when calling ReadSymbolData and
return false instead.
Change-Id: Ie9acdf811300084f1d5916f4778754f8abca10e0
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3572251
Reviewed-by: Ivan Penkov <ivanpe@chromium.org>
1. Visit DW_TAG_class_type when it's inside DW_TAG_subprogram.
2. Only warn when we can't get the name for the DIE and it has DW_AT_specification that is not in the specification map.
Bug: 1078932
Change-Id: Id3126aec305658f8f65c01675a8e9e3ea03f3651
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3579855
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Since the introduction of inlined frames, it is not sufficient to check
the stack trace length (== 1) in order to identify context frames.
Updating all location that were depending on this assumption to check
for frame trust level instead.
Change-Id: I98f966889367c2270c268b8e78b67418c89c50f1
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3499020
Reviewed-by: Mark Mentovai <mark@chromium.org>
for Windows x64 stacks.
According to https://reviews.llvm.org/D2474, LLVM does't generate unwind info for leaf function which doesn't touch any callee-saved
registers. According to MSDN, leaf functions can be unwound simply by
simulating a return.
Change-Id: Ic0503e2aca90b0ba5799133ea8439f1b5f2eefda
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3489332
Reviewed-by: Mark Mentovai <mark@chromium.org>
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
This is needed in order to properly detect and highlight frames that
correspond to multiple functions, for example as the result of identical
code folding by the linker.
Bug: google-breakpad:751
Change-Id: I2ee7c147fcff6493c2454383ad5422b38269759a
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3471034
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Inline frames are always of the base-class type (StackFrame). Treating them as derived-class and accessing members was causing heap buffer overflows.
Change-Id: Id4122ab6a31f016933038a1cb63d45d5c38481f5
Reviewed-on: https://chromium-review.googlesource.com/c/breakpad/breakpad/+/3425445
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
