Re: [PATCH v2] tests: set temp variables using 'env' in test function instead of subshell

Jeff King [off-list ref] writes:

On Tue, Mar 18, 2014 at 01:37:39PM -0700, Junio C Hamano wrote:

quoted

quoted

diff --git a/t/t1300-repo-config.sh b/t/t1300-repo-config.sh
index c9c426c..3e3f77b 100755
--- a/t/t1300-repo-config.sh
+++ b/t/t1300-repo-config.sh

@@ -974,24 +974,15 @@ test_expect_success SYMLINKS 'symlinked configuration' '
 '

 test_expect_success 'nonexistent configuration' '
-	(
-		GIT_CONFIG=doesnotexist &&
-		export GIT_CONFIG &&
-		test_must_fail git config --list &&
-		test_must_fail git config test.xyzzy
-	)
+	test_must_fail env GIT_CONFIG=doesnotexist git config --list &&
+	test_must_fail env GIT_CONFIG=doesnotexist git config test.xyzzy
 '

Isn't GIT_CONFIG here another way of saying:

  test_must_fail git config -f doesnotexist --list

Perhaps that is shorter and more readable still (and there are a few
similar cases in this patch.

Surely, but are we assuming that "git config" correctly honors the
equivalence between GIT_CONFIG=file and -f file, or is that also
something we are testing in these tests?

On Tue, Mar 18, 2014 at 03:16:27PM -0700, Junio C Hamano wrote:

quoted

Isn't GIT_CONFIG here another way of saying:

  test_must_fail git config -f doesnotexist --list

Perhaps that is shorter and more readable still (and there are a few
similar cases in this patch.

Surely, but are we assuming that "git config" correctly honors the
equivalence between GIT_CONFIG=file and -f file, or is that also
something we are testing in these tests?

I think we can assume that they are equivalent, and it is not worth
testing (and they are equivalent in code since 270a344 (config: stop
using config_exclusive_filename, 2012-02-16).

My recollection is that GIT_CONFIG mostly exists as a historical
footnote. Recall that at one time it affected all commands, but that had
many problems and was done away with in dc87183 (Only use GIT_CONFIG in
"git config", not other programs, 2008-06-30). I think we left it in
place for git-config mostly for backward compatibility, but I didn't see
that point explicitly addressed in the list discussion (the main issue
was that setting it for things besides "git config" is a bad idea, as it
suppresses ~/.gitconfig).

-Peff