On Mon, Aug 20, 2012 at 12:28 PM, Jonathan "Duke" Leto <[email protected]> wrote: > Howdy, > >> It is undefined which thread will receive the signal (I asked Leont), >> but with "some"(TM) hopefully the wrong thread will catch it. >> >> With the tsan tool I could repro it always. > > What about teaching the Parrot test suite to use tsan, if it is available? > > It seems like an extremely useful tool to weed out bugs like this. Obviously, > it would be an optional dependency that Configure.pl would detect, and then > our > tests could run certain code with tsan. > > Reasonable or insane?
Reasonable, but there are a lot of races detected. But I would rather wait for clang3.2 where -fthread-sanitizer is builtin, and you can properly skip certain races to be reported. -- Reini Urban http://cpanel.net/ http://www.perl-compiler.org/ _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
