Skip to content

fix(macos): Fix thread safety in NSException capture #7672

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
philprime wants to merge 3 commits into from
Closed

fix(macos): Fix thread safety in NSException capture #7672

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0814e17
fix(macos): Fix thread safety in NSException capture
philprime Mar 11, 2026
3e4252e
docs: Add changelog entry for NSException thread safety fix
philprime Mar 11, 2026
adfb2ba
Merge remote-tracking branch 'origin/main' into philprime/atomic-cras…
philprime Mar 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@

### Fixes

- Fix thread safety in macOS NSException capture to prevent duplicate exception reports (#7672)
- Capture transactions that finish during the launch profiling window instead of silently discarding them (#7602)
- Capture instance and class method `[NSApplication _crashOnException]` exceptions (#7510)

Expand Down
11 changes: 7 additions & 4 deletions Sources/Sentry/SentryNSExceptionCaptureHelper.m
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,28 +5,31 @@
# import "SentryCrash.h"
# import "SentryNSExceptionCaptureHelper.h"
# import "SentrySwift.h"
# import <stdatomic.h>

@implementation SentryNSExceptionCaptureHelper

static BOOL _insideReportException = NO;
// Thread-safe flag to prevent duplicate exception captures when
// _crashOnException: is called from within reportException:
static _Atomic BOOL _insideReportException = NO;

+ (void)reportException:(NSException *)exception
{
_insideReportException = YES;
atomic_store(&_insideReportException, YES);
[self captureException:exception];
}

+ (void)reportExceptionDidFinish
{
_insideReportException = NO;
atomic_store(&_insideReportException, NO);
}

+ (void)crashOnException:(NSException *)exception
{
// When called from within reportException: (i.e., [super reportException:] internally
// dispatches to _crashOnException: when NSApplicationCrashOnExceptions is YES),
// the exception was already captured, so skip to avoid duplicate reports.
if (!_insideReportException) {
if (!atomic_load(&_insideReportException)) {
Copy link
Copy Markdown
Member

@philipphofmann philipphofmann Mar 12, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

h: atomic_store/atomic_load are called in the _crashOnException: path, which runs right before abort(). C11 atomics are not on POSIX's async-signal-safe functions list.

Klaus believes it's OK to call these methods here, but I don't trust him:

  • ATOMIC_BOOL_LOCK_FREE should be 2 (always lock-free) on Apple arm64/x86_64, since _Bool is 8 bits and Clang's MaxAtomicInlineWidth is 128 bits on both architectures (Clang source, LLVM Atomics Guide)
  • If lock-free, these should compile to single CPU instructions (MOV/XCHG on x86, LDAR/STLR on ARM) — no locks, no library calls

Did double-check that it's safe to call C11 atomics in this context? If you already verified this, great — just want to make sure. We could also double-check with @supervacuus.

[self captureException:exception];
}
}
Expand Down
Loading