lberki
lberki
/cc @laszlocsomor Well... the problem is that what RPATH you use is actually important no matter where you use the binary. So I'd rather we take a step back and...
@Wyverald what do you think about this alternative proposal? I essentially took your "use a Thread" change and ran with it -- it turned out to be much simpler than...
This one still has one deadlock (that I know of, anyway) which is shared by every alternative: if the worker thread gets hit by a SIGINT in `delegateEnvQueue.take()` in `signalForFreshEnv()`,...
This alternative is free of any deadlocks I know of. The reason why I stopped using `SynchronousQueue` is that we'd need two of them and I think that doesn't work:...
> > * Is `SkyKeyComputeState.close()` called by Skyframe in any other circumstance than memory pressure? I'm not sure if it would make a difference since it's idempotent and I like...
Update: I tried to use `AtomicReference` for `state.workerThread` and failed at the last step -- it turns out that with `test_keep_going_weird_deadlock()`, we have a case where a single state instance...
Discarding in favor of #22215 and #22100.
Related: https://github.com/bazelbuild/bazel/issues/15029
#13281 is different, I think: it's not about runfiles, but about unnecessarily building things twice.
Is it possible to not audit every call site of the `PathFragment` absolutizaton logic but to add an optional argument that says "absolutize as if the path fragment was implemented...