Valery Mironov
Valery Mironov
@kroeckx First of all, even if it's theoretic only issue, I think it should be fixed if it doesn't affect perf (it doesn't). Secondly, how you can guarantee that all...
@nhorman It's true, but as I wrote it's different bug. This issue about memory order, not about incorrect usage ref print count > Another bug that I noticed :(
> Would that make sense? Sounds like good solution to me. Can you contribute it? Probably not now :(
std::atomic also doesn't work btw
> Interestingly this passes Abseil's tests but breaks almost our entire monorepo in certain cases where operator is defined. I haven't had a chance to dig into why, and I'm...
Ok, thanks for explaining
@chriskohlhoff is any chance you will looking on this? For now it's just leak when this move assign called
Я потыкал из любопытства немного, из интересного мне ответили, и сказали что в принципе можно заюзать dom builder чтобы сразу в binary json строить (проблема что патчить нужно) https://github.com/simdjson/simdjson/discussions/2279#discussioncomment-10988814 Еще...
@t8m I think load used to compare with one, not zero. To understand that we're last **unique** user. For an example if some object shared between threads it cannot be...
> > For an example here, we probably need acquire when nref == 1 > > That code is run inside a mutex so IMO not relevant. but it's different...