I finally figured out what went wrong and it's almost embarrassing: The crash occurred specifically on single core machines as I was allocating a worker thread for each available core, but always subtracted 1 for the main thread. On single core machines this meant that no workers were allocated causing V8 to crash on the as soon as garbage collection was triggered.
On Saturday, 3 May 2025 at 11:46:11 UTC+2 Kasper Isager Dalsgarð wrote: > Another colleague is now hitting this on a Linux machine running under > KVM. I'm still no closer to figuring out where things go wrong; any input > would be greatly appreciated. > > On Thursday, 15 August 2024 at 14:00:29 UTC+2 Kasper Isager Dalsgarð wrote: > >> After several attempts at debugging this, I *think* I've narrowed it >> down to the some sort of interaction with QEMU as it consistently happens >> when running V8 on an emulated Android device on both 32- and 64-bit >> architectures. >> >> On Thursday 7 March 2024 at 09:48:23 UTC+1 Kasper Isager Dalsgarð wrote: >> >>> A colleague of mine briefly hit the same assertion on Android arm64-v8a, >>> but has since been unable to reproduce it. I still haven't been able to >>> narrow down what triggers it. >>> >>> On Thursday 22 February 2024 at 12:07:22 UTC+1 Kasper Isager Dalsgarð >>> wrote: >>> >>>> I'm consistently observing a crash on Android armeabi-v7a during >>>> scavenge collection triggered by `String::NewFromUtf8(Isolate *, const >>>> char >>>> *)`, though I don't suspect that the exact call that triggered the >>>> scavenge >>>> collection relates to the issue. For reference, this is the failing debug >>>> check I'm hitting: >>>> >>>> ``` >>>> # >>>> # Fatal error in ../../../v8/src/heap/scavenger.cc, line 393 >>>> # Debug check failed: copied_list.IsEmpty(). >>>> # >>>> ``` >>>> >>>> I'm having a really hard time debugging why this happens and the fact >>>> that it doesn't happen on arm64-v8a has me even more puzzled. I've read >>>> through https://groups.google.com/g/v8-users/c/sdU232XmyOw where the >>>> same debug check failed, but the was caused by a race condition whereas >>>> the >>>> code I'm running uses only a single thread. >>>> >>>> Which APIs could cause that check to fail? I've tried removing all my >>>> uses of weak persistent handles, for example, but that doesn't make a >>>> difference. Could it my `Platform` implementation that breaks some >>>> invariants that I'm not aware of? >>>> >>>> All of the code that interacts with V8 is available here by the way: >>>> https://github.com/holepunchto/libjs. I of course don't expect anyone >>>> to read through all of that, I'll provide excerpts and summaries of >>>> portions that might be relevant to the issue. >>>> >>>> Any help would be much appreciated, thanks! >>>> >>> -- -- v8-users mailing list v8-users@googlegroups.com 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 v8-users+unsubscr...@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/v8-users/2b3123c0-800d-4c99-be31-6f26d09f8df8n%40googlegroups.com.
