Looking at the current state of the code, I struggle to see how that could
happen. Did you try reproducing on ToT as well?

Also, can you file a bug at https://issues.chromium.org/issues/new with the
component set to "Chromium > Blink > JavaScript > GarbageCollection"?
Please include the stack trace, your GN args, the repro, and any other
relevant information you think could be useful.
If you can have a standalone repro that doesn't require your git repo, that
would be preferable.

On Sat, Aug 9, 2025 at 8:44 PM Kasper Isager Dalsgarð <
[email protected]> wrote:

> Using V8 v13.8.258.18 I'm consistently hitting the following DCHECK
> failure during heap garbage collection:
>
> ```
> #
> # Fatal error in ../../../v8/src/heap/memory-reducer.cc, line 229
> # Debug check failed: 0 < delay_ms (0 vs. 0).
> #
> ```
>
> This is the stack trace at the time of the failure:
>
> ```
> * thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BREAKPOINT
> (code=1, subcode=0x1076866cc)
>   * frame #0: 0x00000001076866cc libjs.dylib`v8::base::OS::Abort()
> [inlined] v8::base::OS::Abort()::$_0::operator()(this=<unavailable>) const
> at platform-posix.cc:729:7 [opt]
>     frame #1: 0x00000001076866cc libjs.dylib`v8::base::OS::Abort() at
> platform-posix.cc:729:7 [opt]
>     frame #2: 0x00000001076713c0 libjs.dylib`V8_Fatal(file=<unavailable>,
> line=<unavailable>, format=<unavailable>) at logging.cc:215:3 [opt]
>     frame #3: 0x0000000107670cb8 libjs.dylib`v8::base::(anonymous
> namespace)::DefaultDcheckHandler(file=<unavailable>, line=<unavailable>,
> message=<unavailable>) at logging.cc:59:3 [opt]
>     frame #4: 0x00000001063feff0
> libjs.dylib`v8::internal::MemoryReducer::ScheduleTimer(this=0x00006070000014b0,
> delay_ms=0) at memory-reducer.cc:229:3 [opt]
>     frame #5: 0x00000001063ff1a4
> libjs.dylib`v8::internal::MemoryReducer::NotifyMarkCompact(this=0x00006070000014b0,
> committed_memory_before=<unavailable>) at memory-reducer.cc:123:5 [opt]
>     frame #6: 0x0000000106337084
> libjs.dylib`v8::internal::Heap::CollectGarbage(v8::internal::AllocationSpace,
> v8::internal::GarbageCollectionReason,
> v8::GCCallbackFlags)::$_0::operator()(this=0x000000016fdfe318) const at
> heap.cc:1714:26 [opt]
>     frame #7: 0x0000000106336ab0 libjs.dylib`void
> heap::base::Stack::SetMarkerAndCallbackImpl<v8::internal::Heap::CollectGarbage(v8::internal::AllocationSpace,
> v8::internal::GarbageCollectionReason,
> v8::GCCallbackFlags)::$_0>(stack=0x000061d000001ba0,
> argument=<unavailable>, stack_end=<unavailable>) at stack.h:185:5 [opt]
>     frame #8: 0x000000010781febc
> libjs.dylib`PushAllRegistersAndIterateStack + 40
>     frame #9: 0x0000000106311e88
> libjs.dylib`v8::internal::Heap::CollectGarbage(v8::internal::AllocationSpace,
> v8::internal::GarbageCollectionReason, v8::GCCallbackFlags) [inlined] void
> heap::base::Stack::SetMarkerIfNeededAndCallback<v8::internal::Heap::CollectGarbage(v8::internal::AllocationSpace,
> v8::internal::GarbageCollectionReason,
> v8::GCCallbackFlags)::$_0>(this=<unavailable>, callback=(unnamed class) @
> 0x000000016fdfe318) at stack.h:81:7 [opt]
>     frame #10: 0x0000000106311e80
> libjs.dylib`v8::internal::Heap::CollectGarbage(this=0x0000633000010938,
> space=<unavailable>, gc_reason=kFinalizeMarkingViaTask,
> gc_callback_flags=<unavailable>) at heap.cc:1655:11 [opt]
>     frame #11: 0x000000010631f51c
> libjs.dylib`v8::internal::Heap::FinalizeIncrementalMarkingAtomically(v8::internal::GarbageCollectionReason)
> [inlined] v8::internal::Heap::CollectAllGarbage(this=0x0000633000010938,
> gc_flags=<unavailable>, gc_reason=kFinalizeMarkingViaTask,
> gc_callback_flags=<unavailable>) at heap.cc:1266:3 [opt]
>     frame #12: 0x000000010631f50c
> libjs.dylib`v8::internal::Heap::FinalizeIncrementalMarkingAtomically(this=0x0000633000010938,
> gc_reason=kFinalizeMarkingViaTask) at heap.cc:3995:3 [opt]
>     frame #13: 0x000000010633bf64
> libjs.dylib`v8::internal::IncrementalMarkingJob::Task::RunInternal(this=0x0000606000016760)
> at incremental-marking-job.cc:135:34 [opt]
> ```
>
> I have a test case at
> https://github.com/holepunchto/libjs/blob/main/test/threads-platform-loop.c
> that reproduces it reliably. It runs an isolate on the main thread that
> does a bunch of large allocations to trigger garbage collection and then a
> separate thread that runs background tasks posted to the embedder platform
> implementation.
>
> It doesn't seem to cause any observable behavior in production builds
> where the corresponding DCHECK is disabled, but I still can't help but
> worry that an invariant is being violated somewhere in the embedder
> platform implementation.
>
> --
> --
> v8-users mailing list
> [email protected]
> http://groups.google.com/group/v8-users
> ---
> You received this message because you are subscribed to the Google Groups
> "v8-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion visit
> https://groups.google.com/d/msgid/v8-users/81831323-681c-454b-816c-66a1f3772b99n%40googlegroups.com
> <https://groups.google.com/d/msgid/v8-users/81831323-681c-454b-816c-66a1f3772b99n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>

-- 
-- 
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/v8-users/CAJFGqfNWp6Gt7%3DWtXdiMNwPChF6T70pWYvUhOZhGKJhuxT9JpA%40mail.gmail.com.

Reply via email to