Re: [PATCH 1/3] t0301: fixes for windows compatibility

Carlo Arenas [off-list ref] writes:

On Sun, Sep 12, 2021 at 10:34 PM Bagas Sanjaya [off-list ref] wrote:

quoted

On 13/09/21 03.28, Carlo Marcelo Arenas Belón wrote:

quoted

test -S is not able to detect that a file is a socket, so use
test -f instead.

Isn't test -f just check for socket as regular file?

and that is exactly how they look; ironically a -f check in Linux
fails for sockets so maybe better to do -e?

an empty file with nothing that indicates in Windows Explorer or a
stat call (from WSL or git bash), that they are anything else.

It actually is a quite attractive idea to use "-e", or even more
preferrably, test_path_exists.  For example:

@@ -31,7 +42,7 @@ helper_test cache
 
 test_expect_success "use custom XDG_CACHE_HOME if set and default sockets are not created" '
 	test_when_finished "git credential-cache exit" &&
-	test -S "$XDG_CACHE_HOME/git/credential/socket" &&
+	test $FLAG "$XDG_CACHE_HOME/git/credential/socket" &&
 	test_path_is_missing "$HOME/.git-credential-cache/socket" &&
 	test_path_is_missing "$HOME/.cache/git/credential/socket"
 '

test_path_exists contrasts better with the two test_path_is_missing
and explains what is being tested better.  Before this part, we have
run some "git credential" test, and there are three possible places
that the socket may appear (XDG, HOME/.git-credential-cache/ and
HOME/.cache/git/credential/), and we want to make sure only one of
them gets it.

One possible downside is that it makes us rely more on our knowledge
that we communicate via unix-domain socket (i.e. what the "socket"
the test is checking is).  By assuming that a mere presence of some
filesystem entity at the inspected path is OK, we may not notice a
breakage that creates a regular file or a directory there by mistake,
yet successfully carry out the credential tests.  It may even be a
good thing, if future ourselves have somehow found out how to use a
regular file or a directory for IPC instead of using a socket ;-).

Thanks.