On Thu, Nov 22, 2018 at 02:17:01AM -0800, Carlo Arenas wrote: > Peff, could you elaborate on your "load testing" setup? which could > give us any hints > on what to look for?, FWIW I hadn't been able to reproduce the problem anywhere > else (and not for a lack of trying) The script I use is at: https://github.com/peff/git/blob/meta/stress which you invoke like "/path/to/stress t5562" from the top-level of a git.git checkout. It basically just runs a loop of twice as many simultaneous invocations of the test script as you have CPUs, and waits for one to fail. The load created by all of the runs tends to flush out timing effects after a while. It fails for me on t5562 within 30 seconds or so (but note that in this particular case it sometimes takes a while to produce the final output because invoke-with-content-length misses the expected SIGCLD and sleeps the full 60 seconds). You'll probably need to tweak the variables at the top of the script for your system. > PS. upstreaming the PERL_PATH fix is likely to be good to do soonish > as I presume at least all BSD might be affected, let me know if you > would rather me do that instead as I suspect we might be deadlocked > otherwise ;) Yeah, the $PERL_PATH thing is totally orthogonal, and should graduate separately. -Peff
