On Tue, Nov 29 2022, Alison Winters wrote: > On 2022-11-29 10:38, Junio C Hamano wrote: >> I did not try to figure out the reason but the topic with its tests >> seem to break in 'seen' the linux-cmake-ctest CI job. >> >> https://github.com/git/git/actions/runs/3570230611/jobs/6001013549 >> >> but the same test does not break under usual "make test". > > I cannot speak for how the changes of the ab/cmake-nix-and-ci impact > the aw/complete-case-insensitive branch, but the failure seems to be > pointing to a test that I have since changed in the v2 patch, on the > suggestion of Szeder Gabor. The variable is no longer exported and > the script is no longer sourced a second time. I don't know if those > v2 changes would change the results of this test, but they might be a > starting point for the other people CC:ed here to consider. The breakage is due to a semantic conflict between the two, which is solved thusly: diff --git a/t/t9902-completion.sh b/t/t9902-completion.sh index d45d820da28..2aa128075ca 100755 --- a/t/t9902-completion.sh +++ b/t/t9902-completion.sh @@ -2261,7 +2261,7 @@ test_expect_success 'checkout does not match ref names of a different case' ' test_expect_success 'checkout matches case insensitively with GIT_COMPLETION_IGNORE_CASE' ' ( - . "$GIT_BUILD_DIR/contrib/completion/git-completion.bash" && + . "$GIT_SOURCE_DIR/contrib/completion/git-completion.bash" && GIT_COMPLETION_IGNORE_CASE=1 && export GIT_COMPLETION_IGNORE_CASE && test_completion "git checkout M" <<-\EOF main Z @@ -2279,7 +2279,7 @@ test_expect_success 'checkout completes pseudo refs' ' test_expect_success 'checkout completes pseudo refs case insensitively with GIT_COMPLETION_IGNORE_CASE' ' ( - . "$GIT_BUILD_DIR/contrib/completion/git-completion.bash" && + . "$GIT_SOURCE_DIR/contrib/completion/git-completion.bash" && GIT_COMPLETION_IGNORE_CASE=1 && export GIT_COMPLETION_IGNORE_CASE && test_completion "git checkout h" <<-\EOF HEAD Z Junio: I don't think this warrants kicking out the cmake topic. This sort of problem is just going to occur while something like that is in-flight, or Alison would have otherwise presumably seen it as a CI failure if based off "master". With hindsight I could have made e74e51a9154 (cmake & test-lib.sh: add a $GIT_SOURCE_DIR variable, 2022-11-03) do the migration in two phases, but I didn't expect another topic to rely on the relatively obscure bits that were being copied to the $GIT_BUILD_DIR, we weren't exactly drowning in git-completion.bash patches...
