On Thu, Apr 9, 2020 at 11:39 AM Davide Italiano <ditali...@apple.com> wrote:
> > > > On Apr 9, 2020, at 11:05, Jonas Devlieghere via lldb-commits < > lldb-commits@lists.llvm.org> wrote: > > > > > > Author: Jonas Devlieghere > > Date: 2020-04-09T11:03:24-07:00 > > New Revision: 680082a408dd2df7410d77696100eac8ce9d5530 > > > > URL: > https://github.com/llvm/llvm-project/commit/680082a408dd2df7410d77696100eac8ce9d5530 > > DIFF: > https://github.com/llvm/llvm-project/commit/680082a408dd2df7410d77696100eac8ce9d5530.diff > > > > LOG: [lldb/Reproducers] Add a small artificial delay before exiting > > > > Add a small artificial delay in replay mode before exiting to ensure > > that all asynchronous events have completed. This should reduce the > > level of replay flakiness on some of the slower bots. > > > > Added: > > > > > > Modified: > > lldb/source/Utility/ReproducerInstrumentation.cpp > > > > Removed: > > > > > > > > > ################################################################################ > > diff --git a/lldb/source/Utility/ReproducerInstrumentation.cpp > b/lldb/source/Utility/ReproducerInstrumentation.cpp > > index 4c32d9407830..3bf81286bbe9 100644 > > --- a/lldb/source/Utility/ReproducerInstrumentation.cpp > > +++ b/lldb/source/Utility/ReproducerInstrumentation.cpp > > @@ -8,6 +8,7 @@ > > > > #include "lldb/Utility/ReproducerInstrumentation.h" > > #include "lldb/Utility/Reproducer.h" > > +#include <thread> > > #include <stdio.h> > > #include <stdlib.h> > > > > @@ -94,6 +95,10 @@ bool Registry::Replay(llvm::StringRef buffer) { > > GetReplayer(id)->operator()(deserializer); > > } > > > > + // Add a small artificial delay to ensure that all asynchronous > events have > > + // completed before we exit. > > + std::this_thread::sleep_for (std::chrono::milliseconds(100)); > > + > > return true; > > } > > > > I understand this is (unfortunately) sorely needed, but I’m somehow > skeptical of random delays. > You probably thought about this and dropped the idea on the floor — but, > is there a way to make the sync more explicit? > If you can’t, at least you can detect not everything is in order and > sleep, for, e.g. another 100 milliseconds in a retry-loop? > Yeah, it's a hack, and I'm not really happy with it. The last replayed API call can be literaly anything, so you don't know what you might have to wait for. I can't think of a good way to do the synchronization without breaking layering/abstractions. The issue on GreenDragon was that we exited before the stop event came in, and while we could probably synchronize that, it would be even more ad-hoc than this... > > — > Davide
_______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
