On Tue, 2016-05-03 at 22:34 +0100, Ramsay Jones wrote: > Hi David, > > Test t7900.5 fails for me, thus: > > $ ./t7900-index-helper.sh -i -v -x -d > > ... > > + test -S .git/index-helper.sock > + git status > On branch master > Untracked files: > (use "git add <file>..." to include in what will be committed) > > err > > nothing added to commit but untracked files present (use "git add" > to track) > + test -S .git/index-helper.sock > + grep -q . err > error: last command exited with $?=1 > not ok 5 - index-helper autorun works > # > # rm -f .git/index-helper.sock && > # git status && > # test_path_is_missing .git/index-helper.sock && > # test_config indexhelper.autorun true && > # git status && > # test -S .git/index-helper.sock && > # git status 2>err && > # test -S .git/index-helper.sock && > # ! grep -q . err && > # git index-helper --kill && > # test_config indexhelper.autorun false && > # git status && > # test_path_is_missing .git/index-helper.sock > # > $ cd trash\ directory.t7900-index-helper/ > $ ls > err x.t > $ cat err > warning: We requested watchman support from index-helper, but it > doesn't support it. Please use a version of git index-helper with > watchman support. > $ > > [Yes, that is one long line in err!] OK, the failure here is due to a bad check in the code -- it's intended that index-helper work without watchman. Will re-roll with a fix. > [At least, this is one of the failures, I have also seen git status > failing > with a SIGPIPE.] I'll add in a sig_ign on that. > Note that I do not have the watchman libraries etc., so USE_WATCHMAN > is > not defined. Note also, that I had an instance of git-index-helper > still > running after the test failure (which I kill-ed). I missed a test_when_finished (note that in the case of test failures in debug mode, test_when_finished statements are not executed on failing tests, so you would still see an index-helper in that case; since test failures are not expected, that shouldn't be a problem). Will fix, thanks. > I haven't spent any time debuging this, but some questions spring to > mind: > > - can index-helper be used without watchman support? Yes (once I re-roll with the bug fix). > - why is index-helper requesting watchman support, when it was > built without USE_WATCHMAN being defined? It's not; the code that checks for that request is mistakenly detecting that it is. > - why is read-cache.o exporting hte verify_index and > write_watchman_ext symbols? verify_index should be used in index-helper but that somehow got lost in v3. Will fix on reroll. write_watchman_ext is exported because index-helper currently uses it. > - is index-helper any use/help without watchman support? Yes for large indexes -- see the measurements in patch v3 (which is earlier in the series than watchman is introduced). > - is '! grep -q . err' meant to determine if the err file is > empty (ie git status did not issue an error message)? > [if yes, maybe 'test_must_be_empty err &&' would read better!] Will fix, thanks. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html