Slava Egorov
Slava Egorov
@brianquinlan What I see is that baseline seems to have regressed at some point. If I measure timings using 810301045d700c9f40406c2c5a46d82ac487ffb9 as baseline I see ~1.5s both with and without my...
It seems that the actual time is spent prior to `main` in compilation. When I benchmarked I simply build `dart` target - which does not rebuild `kernel-service.dart.snapshot`. If `dart` fails...
> Awesome! Let me know if I can do anything to help you land this. I am a bit busy this week (and the start of the next week), but...
The first half of the problem is now fixed by 5a32d8b. Opening a pipe like that no longer causes unrelated operations to hang. The program overall still hangs however -...
(Also if you open enough pipes you will still saturate available thread pool for IO tasks - so the problem is only kinda fixed)
Well, it is somewhat mitigated... If the code opens >=32 pipes for reading things will get stuck again. So it would be nice to make a patch to fix the...
There are some things I like about this proposal (e.g. every shared field automatically gets atomic APIs) and some things I don't like. Being shared or not-shared is an attribute...
Where are we currently standing on this issue? @bwilkerson @DanTup
I think this is another manifestation of https://issues.chromium.org/issues/340777103, there is no real leak - but the reclamation of objects is so delayed that it looks like a leak. `_PlaceholderPainter.paint` creates...