On 10/29/2014 08:36 PM, Bernhard Voelker wrote: > On 10/29/2014 03:15 PM, Pádraig Brady wrote: >> * tests/tail-2/inotify-rotate.sh: > > Wow, what coincidence: exactly today that test started failing > for the first time in openSUSE's Build Service: > > https://build.opensuse.org/package/show/Base:System/coreutils-testsuite > https://build.opensuse.org/build/Base:System/openSUSE_Factory/x86_64/coreutils-testsuite/_log > > Somehow I have the impression that this is some fallout of the > recent update of glibc to 2.20. > > Have a nice day, > Berny > > > [ 736s] FAIL: tests/tail-2/inotify-rotate > [ 736s] =================================
> [ 736s] + for i in '$(seq 50)' > [ 736s] + echo 1 > [ 736s] 1 > [ 736s] + rm -rf k x out > [ 736s] + pid=24570 > [ 736s] + sleep .1 > [ 736s] + timeout 40 tail -F k > [ 736s] + echo b > [ 736s] + grep_timeout b out > [ 736s] + local j > [ 736s] ++ seq 150 > [ 736s] + for j in '$(seq 150)' > [ 736s] + grep b out > [ 736s] + return 0 > [ 736s] + : > [ 736s] + grep b out > [ 736s] + break > [ 736s] + mv x k > [ 736s] + grep_timeout tail: out > [ 736s] + local j > [ 736s] ++ seq 150 > [ 736s] + for j in '$(seq 150)' > [ 736s] + grep tail: out > [ 736s] + return 0 > [ 736s] + echo ok > [ 736s] + found=0 > [ 736s] + grep_timeout ok out > [ 736s] + for j in '$(seq 150)' > [ 736s] + grep ok out > [ 736s] + sleep 0.1 > [ 736s] + return 1 > [ 736s] + kill 24570 > [ 736s] ./tests/tail-2/inotify-rotate.sh: line 65: kill: (24570) - No such > process I was wonderomg did the timeout 40 kill the tail process before it could propagate the "ok" I.E. that something recent was slowing down this test causing the false positives? Though the the above happened on the first time through the loop so that's not likely. In the hydra case, tail did output but not fast enough so it seemed to be a genuine race. Also it seems glibc 2.19 is used in the hydra case: http://hydra.nixos.org/build/16546517#tabs-runtime-deps BTW I must look into why the verbose logs aren't output by hydra even though VERBOSE=yes is set in the hydra scripts. thanks, Pádraig.
