Constantine Peresypkin

> My fix landed in v8 today -- https://chromium-review.googlesource.com/c/v8/v8/+/2384166 Reverted, it seems.

@laverdet are you ok if I backport the patch to v8 7.x ? Integrating 8.x is tricky in some places. :)

@laverdet nope, a PR on plv8 which uses 7.x releases for now

@laverdet It seems like there's still something strange going on. Example code: https://gist.github.com/pkit/427d3ac691882782c61e6ceec490348c Running it like: ``` $ ./test1 'for(;;){}' timeout = 0.500s timed out elapsed: 0.500s ``` ``` $...

@laverdet yup, tested that too, will never terminate without a patch. But cannot find where the 10 sec stuff comes from. It looks like the interrupts are checked each 1...

@laverdet reducing chunk size to 256 (or even 128) and breaking the multiplier loop doesn't slow down the conversion from my benchmarks. But allows for ~ 100ms timer resolution.

Nope, not at all, I just wanted to fix at least the known uninterruptable stuff. For now I can just recompile custom v8 when needed.

Hmm, most of these lead to OOM handler invocation which will either call `abort()` or return to V8 which will call `abort()` anyway. In `plv8` I have a stack manipulation...

Yup, `setsigjump` and friends. I don't see anything non-trivial in the v8 code there though. And they do allocate things on stack for all the contexts. I can kill the...

Hmm, let's consider the following code: ``` void SomeFunction(Isolate *isolate) { Isolate::Scope scope(isolate); HandleScope handle_scope(isolate); Local context = GetSomeContext(); Context::Scope context_scope(context); .... isolate->Exit(); isolate->Destroy(); } ``` What would happen with...