The isolate can indeed be re-entered and, in that case, a counter is increased. The isolate is only exited when the counter becomes zero. This is indeed a different behaviour from what the documentation <https://source.chromium.org/chromium/chromium/src/+/main:v8/include/v8-locker.h;drc=65034fdf545d2f937dd05f1aa535c39ba21ccac5;l=69> specifies for the Locker/Unlocker.
Do you have a scenario where it is not possible or convenient to properly exit the isolate (using the right number of Exit calls) before using the Unlocker? If thread A has entered the isolate (N times) and then uses the Unlocker without properly exiting (N times), then it can still use the isolate. If at the same time thread B uses the Locker and enters the isolate, we end up with two threads using the isolate simultaneously. On Tuesday, January 16, 2024 at 4:56:51 PM UTC+1 me...@redis.com wrote: > Nikos I am still confuse. > > The documentation also state that reenter the isolate is allowed: > https://source.chromium.org/chromium/chromium/src/+/main:v8/include/v8-isolate.h;l=731 > So before creating an unlocker we must exit the isolate the number of > times we entered it (and then re-enter it again the same number of times > after the unlocker is destroyed)? > > Or is there some mechanism that handles it when we exit/enter the isolate? > > On Tuesday 16 January 2024 at 13:58:19 UTC+2 Meir Shpilraien wrote: > >> Thanks Nikos, >> >> You are right, I missed it in our documentation. Will fix our code. >> >> Thank you for the quick response. >> >> On Tuesday 16 January 2024 at 13:46:19 UTC+2 niko...@google.com wrote: >> >>> Hi Meir, >>> >>> You are correct about the purpose of the CHECK that I introduced in CL >>> 5173257 <https://chromium-review.googlesource.com/c/v8/v8/+/5173257>. >>> >>> The documentation >>> <https://source.chromium.org/chromium/chromium/src/+/main:v8/include/v8-locker.h;drc=65034fdf545d2f937dd05f1aa535c39ba21ccac5;l=45> >>> >>> of the Locker/Unlocker API specifies that a thread has to exit the isolate >>> before using v8::Unlocker and enter it again afterwards. >>> >>> Without seeing your code, I cannot be sure that this is the issue but >>> from what you write I suspect that it is. It should also be very easy to >>> fix. >>> >>> See here for a similar fix that had to be made in chromium: CL 5168625 >>> <https://chromium-review.googlesource.com/c/chromium/src/+/5168625>. >>> >>> Best regards, >>> Nikos. >>> >>> On Tuesday, January 16, 2024 at 12:18:22 PM UTC+1 me...@redis.com wrote: >>> >>> Hey, >>> >>> We recently encounter an issue in which our code, when run on V8 main >>> branch, crashes with this fatal error: >>> # >>> # Fatal error in ../../src/execution/isolate.cc, line 4996 >>> # Check failed: current_thread_id_ == thread_id (<unprintable> vs. >>> <unprintable>). >>> # >>> # >>> # >>> >>> I saw that this check was added recently on this PR: >>> https://chromium-review.googlesource.com/c/v8/v8/+/5173257 >>> >>> Its purpose is to verify a correct use of entering and existing an >>> isolate. But I do not see it is taking into consideration the >>> Locker/Unlocker API. In case the current thread uses Unlocker and then >>> another thread tries to enter the isolate it will failed. >>> >>> Is my analysis correct or am I missing something? >>> Thanks, >>> Meir >>> >>> -- -- v8-dev mailing list v8-dev@googlegroups.com http://groups.google.com/group/v8-dev --- You received this message because you are subscribed to the Google Groups "v8-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-dev+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/v8-dev/1c5f78f7-74f7-4cce-815c-9f670d8145c6n%40googlegroups.com.
