feat: Capturing fatal CPPExceptions via cxa_throw

[philipphofmann]

📜 Description

Use fishhook to swap the __cxa_throw C++ exception handler for getting
stacktraces for unhandled C++ exceptions. This implementation is based
on the code in https://github.com/kstenerud/KSCrash.

💡 Motivation and Context

Fixes GH-4517

💚 How did you test it?

Testing Details macOS sample

Without the feature

https://sentry-sdks.sentry.io/issues/6590753526/events/d1bb9784083e40a8abfdb28ac2b9c34e/?project=5428557

With the feature

Invalid Argument

sentry-cocoa/Samples/macOS-Swift/Shared/CppSample.cpp

Lines 4 to 8 in e70c3aa

	void
	internalFunction(void)
	{
	throw std::invalid_argument("Internal function must not be called.");
	}

https://sentry-sdks.sentry.io/issues/6545237791/events/a393feb8d8b2401c8d6cedca5ce181a4/?project=5428557

rethrowNoActiveCPPException

sentry-cocoa/Samples/macOS-Swift/Shared/CppSample.cpp

Lines 22 to 28 in e70c3aa

	void
	Sentry::CppSample::rethrowNoActiveCPPException(void)
	{
	// Rethrowing an exception when there is no active exception will lead to std::terminate being
	// called.
	throw;
	}

https://sentry-sdks.sentry.io/issues/6617441025/events/30bc9db8d17b4dcd80d6473c57a3c36f/?project=5428557

C++ Exception from BG Thread

https://sentry-sdks.sentry.io/issues/6565152294/events/691f845d8dc04305b0cf376e2d21732b/?project=5428557

NoExcept C++ Exception

sentry-cocoa/Samples/macOS-Swift/Shared/CppSample.cpp

Lines 16 to 20 in e70c3aa

	void
	Sentry::CppSample::noExceptCppException() noexcept
	{
	throw std::invalid_argument("Invalid Argument.");
	}

https://sentry-sdks.sentry.io/issues/6565163235/events/d9dc157068994ed2879842dbaf5f477b/?project=5428557

Unhandled C++Exception from testflight iOS-Swift
https://sentry-sdks.sentry.io/issues/6623122757/events/85451edf3f324ad98a6293f1e2cc4b55/?project=5428557

📝 Checklist

You have to check all boxes before merging:

• I added tests to verify the changes.
• No new PII added or SDK only sends newly added PII if sendDefaultPII is enabled.
• I updated the docs if needed.
• I updated the wizard if needed.
• Review from the native team if needed.
• No breaking change or entry added to the changelog.
• No breaking change for hybrid SDKs or communicated to hybrid SDKs.

[github-actions]

	Messages
📖	Do not forget to update Sentry-docs with your feature once the pull request gets approved.

Generated by 🚫 dangerJS against 877a487

[codecov]

Codecov Report

Attention: Patch coverage is 92.85714% with 63 lines in your changes missing coverage. Please review.

Project coverage is 92.892%. Comparing base (9299104) to head (877a487).

Files with missing lines	Patch %	Lines
...Crash/Recording/Tools/SentryCrashCxaThrowSwapper.c	84.210%	39 Missing ⚠️
...ts/SentryCrash/SentryCrashCxaThrowSwapper_Tests.mm	91.011%	16 Missing ⚠️
...rding/Monitors/SentryCrashMonitor_CPPException.cpp	73.076%	7 Missing ⚠️
...ntryCrash/SentryCrashMonitor_CppException_Tests.mm	96.666%	1 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##              main     #5256       +/-   ##
=============================================
- Coverage   92.909%   92.892%   -0.018%     
=============================================
  Files          688       692        +4     
  Lines        86326     87075      +749     
  Branches     31213     31425      +212     
=============================================
+ Hits         80205     80886      +681     
- Misses        6021      6088       +67     
- Partials       100       101        +1     

Files with missing lines	Coverage Δ
Sources/Sentry/SentryCrashIntegration.m	99.186% <100.000%> (+0.027%)	⬆️
Sources/Sentry/include/SentryCompiler.h	100.000% <100.000%> (ø)
...es/SentryCrash/Recording/Tools/SentryCrashMach-O.c	100.000% <100.000%> (ø)
...cording/Tools/SentryCrashPlatformSpecificDefines.h	100.000% <100.000%> (ø)
Sources/Swift/Core/SentryExperimentalOptions.swift	60.000% <100.000%> (+10.000%)	⬆️
...es/Swift/Helper/SentryEnabledFeaturesBuilder.swift	100.000% <100.000%> (ø)
...sts/Helper/SentryEnabledFeaturesBuilderTests.swift	100.000% <100.000%> (ø)
...ions/SentryCrash/SentryCrashIntegrationTests.swift	92.598% <100.000%> (+0.374%)	⬆️
...s/SentryTests/SentryCrash/SentryCrashMach-OTests.m	100.000% <100.000%> (ø)
...ntryCrash/SentryCrashMonitor_CppException_Tests.mm	94.392% <96.666%> (+0.886%)	⬆️
... and 3 more

... and 23 files with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9299104...877a487. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Performance metrics 🚀

	Plain	With Sentry	Diff
Startup time	1224.49 ms	1239.52 ms	15.03 ms
Size	23.76 KiB	825.23 KiB	801.47 KiB

Baseline results on branch: main

Startup times

Revision	Plain	With Sentry	Diff
deeb22c	1233.90 ms	1250.19 ms	16.29 ms
fb7eac9	1236.33 ms	1245.08 ms	8.76 ms
e778bd2	1224.66 ms	1252.16 ms	27.50 ms
a0bb631	1200.55 ms	1214.20 ms	13.65 ms
fa38a2b	1217.92 ms	1235.52 ms	17.60 ms
df50619	1227.12 ms	1243.49 ms	16.37 ms
533c68f	1236.54 ms	1256.68 ms	20.14 ms
b2c9166	1246.86 ms	1255.28 ms	8.42 ms
a6f8b18	1238.54 ms	1265.56 ms	27.02 ms
83acf3e	1240.06 ms	1261.38 ms	21.31 ms

App size

Revision	Plain	With Sentry	Diff
deeb22c	21.58 KiB	612.11 KiB	590.53 KiB
fb7eac9	23.76 KiB	861.22 KiB	837.46 KiB
e778bd2	20.76 KiB	426.15 KiB	405.39 KiB
a0bb631	21.58 KiB	546.19 KiB	524.61 KiB
fa38a2b	21.58 KiB	418.70 KiB	397.12 KiB
df50619	21.58 KiB	671.58 KiB	650.00 KiB
533c68f	21.58 KiB	630.28 KiB	608.70 KiB
b2c9166	21.58 KiB	630.26 KiB	608.68 KiB
a6f8b18	20.76 KiB	431.87 KiB	411.11 KiB
83acf3e	21.58 KiB	539.87 KiB	518.29 KiB

Previous results on branch: fix/cpp-exceptions-v3

Startup times

Revision	Plain	With Sentry	Diff
fa4ec70	1233.90 ms	1253.20 ms	19.30 ms
12a2c05	1213.24 ms	1232.41 ms	19.16 ms
1a55563	1219.78 ms	1245.86 ms	26.08 ms
a87cd4b	1225.20 ms	1245.16 ms	19.96 ms
0143072	1219.62 ms	1252.60 ms	32.98 ms
e102ea7	1239.84 ms	1254.65 ms	14.82 ms

App size

Revision	Plain	With Sentry	Diff
fa4ec70	23.76 KiB	825.20 KiB	801.44 KiB
12a2c05	23.76 KiB	825.23 KiB	801.47 KiB
1a55563	23.76 KiB	825.24 KiB	801.48 KiB
a87cd4b	23.76 KiB	825.30 KiB	801.55 KiB
0143072	23.76 KiB	825.26 KiB	801.49 KiB
e102ea7	23.76 KiB	825.28 KiB	801.51 KiB

[philprime]

I'll wait for @supervacuus feedback on your PR comment before doing the review in case it needs another revision first.

[philipphofmann]

I'll wait for @supervacuus feedback on your PR comment before doing the review in case it needs another revision first.

This is mostly native stuff. I would wait for the review of @supervacuus before adding your review.

[supervacuus]

Quick update: I have already started "investigating" the changes. I hope that my response to your comment makes sense. This conversation should ideally lead to inline comments.

I am still in the process, but most parts look relatively sane. A first questionable impression is the lack of thread synchronization, considering that we access quite a few globals. cxa_throw itself can and will be called from multiple threads simultaneously, which might be less of a problem if we only read. However, there is still a chance that setup and teardown happen concurrently. Did you consider these possible interactions already? It could be fine if you were sure to delineate reads and writes with certainty, but that is rarely the case.

[supervacuus]

This looks great—only minor things. The only unanswered aspect I currently see is that during sentry initialization, when you rebind __cxa_throw with your decorator, another thread could already be throwing and entering your decorator. At the same time, the g_cxa_originals structure is still not finished.

The problem with that is that we must call an "original"; otherwise, we crash. We can either guarantee that the SDK initialization is completed before any other thread can throw, or we create an explicit abort() path that logs the situation, so users aren't confronted with an indecipherable crash in this scenario. And to be clear, this is an edge case, but still a race.

[philipphofmann]

I have no clue why, but the unit tests for Mac Catalyst keep failing. I need to figure out why that is before merging this.