On November 24, 2022 1:41 PM, Ævar Arnfjörð Bjarmason wrote: >[CC-ing Jeff, this is from the "stopwatch timers" topic] > >On Thu, Nov 24 2022, rsbecker@xxxxxxxxxxxxx wrote: > >> While running t2201.7 (and subsequent) for 2.39.0-rc0, I hit a >> condition >> where: > >Don't you mean t0211-trace2-perf.sh? > >> <snip> >> + test_when_finished rm trace.perf actual test_config_global >> + trace2.perfBrief 1 pwd test_config_global trace2.perfTarget >> /home/ituglib/randall/jenkins/.jenkins/workspace/Git_Pipeline/t/trash >> directory.t0211-trace2-perf/trace.perf >> + test-tool trace2 101timer 5 10 3 >> fatal: failed to create thread[0] >> >> was encountered, causing the test to fail. Pointers on resolving this >> would be helpful. > >Apparently we really could use a NO_PTHREADS=Y CI setup, because that's all >that's needed to reproduce this on linux & elsewhere. Running the test with that: > > + test-tool trace2 101timer 5 10 3 > fatal: failed to create thread[0] > error: last command exited with $?=128 > >For that all that's needed is this: > > diff --git a/t/t0211-trace2-perf.sh b/t/t0211-trace2-perf.sh > index 0b3436e8cac..cfba6861322 100755 > --- a/t/t0211-trace2-perf.sh > +++ b/t/t0211-trace2-perf.sh > @@ -203,7 +203,7 @@ test_expect_success 'stopwatch timer test/test1' ' > have_timer_event "main" "timer" "test" "test1" 5 actual > ' > > -test_expect_success 'stopwatch timer test/test2' ' > +test_expect_success PTHREADS 'stopwatch timer test/test2' ' > test_when_finished "rm trace.perf actual" && > test_config_global trace2.perfBrief 1 && > test_config_global trace2.perfTarget "$(pwd)/trace.perf" && > @@ -249,7 +249,7 @@ test_expect_success 'global counter test/test1' ' > have_counter_event "main" "counter" "test" "test1" 15 actual > ' > > -test_expect_success 'global counter test/test2' ' > +test_expect_success PTHREADS 'global counter test/test2' ' > test_when_finished "rm trace.perf actual" && > test_config_global trace2.perfBrief 1 && > test_config_global trace2.perfTarget "$(pwd)/trace.perf" && > >I think that's the correct fix here, and it should go into rc1. I assume we do not have to modify the platform's entry in config.mak.uname, which already has NO_PTHREADS=UnfortunatelyYes. Thanks, Randall
