Junio C Hamano wrote:
Would doing something like this (without any other changes we
discussed so far) help?
t/test-lib.sh | 2 ++
1 file changed, 2 insertions(+)
diff --git c/t/test-lib.sh w/t/test-lib.sh
index af722d383d..92d0db13d7 100644
--- c/t/test-lib.sh
+++ w/t/test-lib.sh
@@ -1577,6 +1577,8 @@ fi
# Use -P to resolve symlinks in our working directory so that the cwd
# in subprocesses like git equals our $PWD (for pathname comparisons).
cd -P "$TRASH_DIRECTORY" || BAIL_OUT "cannot cd -P to \"$TRASH_DIRECTORY\""
+TRASH_DIRECTORY=$(pwd)
+HOME="$TRASH_DIRECTORY"
start_test_output "$0"
That does work as an alternate way to fix the bug that I described.
Forcing the entire test suite to run from a realpath-canonicalized path
seems like a blunt instrument, though. It would tend to mask bugs that
only occur when git runs with a non-canonical path as its working
directory. It could even mask bugs in git's own pathname canonicalization
code.
