On Mon, Aug 23, 2021 at 04:32:39PM -0700, Junio C Hamano wrote:
Mickey Endito [off-list ref] writes:
quoted
The variable D is never defined in test t5582, more severely the test
fails if D is defined by something outside the test suite, so remove
this spurious line.
Wow. Well spotted.
When D is left unset, we end up executing
cd "" && ...
and it explains why nobody noticed the breakage for nearly a year
since c0192df6 (refspec: add support for negative refspecs,
2020-09-30) was written.
Unlike the apparent
copy-and-paste source, this is a more modern script that limits the
chdir inside subshells to avoid moving around in the main flow of
the test, and the fix proposed here looks the most sensible.
'grep " cd $" test-results/*.out' shows that there is a similar case
in 't5323-pack-redundant.sh' as well, in test 'master: pack-redundant
works with no packfile'.
quoted
Signed-off-by: Mickey Endito <redacted>
---
To reproduce a failure do
D=/some/path/which/does/not/exist t/t5582-fetch-negative-refspec.sh
Note: The variable D seems to be a reminiscent similar to t/t5510-fetch.sh,
which defines "D=$(pwd)". If you want to adopt that way, then you have
to code a fix yourself. ;-)
t/t5582-fetch-negative-refspec.sh | 1 -
1 file changed, 1 deletion(-)
diff --git a/t/t5582-fetch-negative-refspec.sh b/t/t5582-fetch-negative-refspec.sh
index e5d2e79ad3..7a80e47c2b 100755
--- a/t/t5582-fetch-negative-refspec.sh
+++ b/t/t5582-fetch-negative-refspec.sh
@@ -105,7 +105,6 @@ test_expect_success "fetch with negative pattern refspec does not expand prefix"
'
test_expect_success "fetch with negative refspec avoids duplicate conflict" '
- cd "$D" &&
(
cd one &&
git branch dups/a &&
--
2.30.2