Re: [PATCH 0/4] tests: use "git xyzzy" form

Thanks.

I looked at the output from:

    $ git grep -e git-'[a-z]' t/t[0-8]???-*.sh t/t9001-*.sh

after applying these patches to the tip of ho/dashless topic, 7e44c93
('git foo' program identifies itself without dash in die() messages,
2008-08-31) to review what you did not change are reasonable.

I won't quote the whole output from the above grep, but there are a few
lines worth commenting on.

t/t5300-pack-object.sh:test_description='git-pack-object
t/t5305-include-tag.sh:test_description='git-pack-object --include-tag'

These might want changing.

t/t5602-clone-remote-exec.sh:	echo "localhost git-upload-pack '\''/path/to/repo'\''" >expected
t/t5602-clone-remote-exec.sh:	GIT_SSH=./not_ssh git clone -u /something/bin/git-upload-pack localhost:/path/to/repo junk
t/t5602-clone-remote-exec.sh:	echo "localhost /something/bin/git-upload-pack '\''/path/to/repo'\''" >expected

These we do want to keep in git-foo form

The remainder of the grep output looked all reasonable; they hit options
like "--git-dir" and "is-inside-git-dir", and pathnames like "git-gui" and
".git/rebase-merge/git-rebase-todo".

Doing this _carefully_ (otherwise the result would be useless) must have
been quite a lot of effort, and what was posted seems to have been done
quite carefully.

The cross-checking I did with "grep" in this message does *not* catch if
you converted something that shouldn't have been converted (e.g. changing
the pathname ".git/rebase-merge/git-rebase-todo" to "git rebase-todo"
would have been a disaster).  It needs real eyeballing of the patches
themselves to catch such an error, and we would need help from people.

Volunteers?  Just please say "I'll go through patch [n/4] line by line"
now to avoid duplicated work with others (more eyeballs looking at the
same thing will catch more errors, but I do not think we have that many
volunteers here), and then after you are done, come back with "They all
are command invocations and conversion is correct" or with "This one is
incorrect and should be kept."

Oh, also, there are many other t9XXX series of tests.  Can somebody
volunteer to do the same for the rest of them?

Puh-leeeeze?

On Wed, Sep 03, 2008 at 01:45:46PM -0700, Junio C Hamano wrote:

Doing this _carefully_ (otherwise the result would be useless) must have
been quite a lot of effort, and what was posted seems to have been done
quite carefully.

Yes, thank you Nanako. After complaining about it to Junio, I expected I
would have to be the one to do it. But procrastination pays off once
again...

The cross-checking I did with "grep" in this message does *not* catch if
you converted something that shouldn't have been converted (e.g. changing
the pathname ".git/rebase-merge/git-rebase-todo" to "git rebase-todo"
would have been a disaster).  It needs real eyeballing of the patches
themselves to catch such an error, and we would need help from people.

When I did this the last time, I recall that most of the errors of that
sort tended to make the test fail loudly. So that is a point in our
favor. However, I agree that eyeballing is a good strategy here.

I just checked 1/4, which looks fine. Made much easier by the use of
--color-words (even easier with -U0, but sometimes the context is
helpful for seeing that it is correct).

In 2/4, we change some instances of git-upload-pack, which you otherwise
mentioned in your "should be left". It's in t5530 and t5602. Otherwise
looks fine.

-Peff

On Wed, Sep 03, 2008 at 04:59:45PM -0400, Jeff King wrote:

I just checked 1/4, which looks fine. Made much easier by the use of
--color-words (even easier with -U0, but sometimes the context is
helpful for seeing that it is correct).

In 2/4, we change some instances of git-upload-pack, which you otherwise
mentioned in your "should be left". It's in t5530 and t5602. Otherwise
looks fine.

And 3/4 and 4/4 look fine to me, too.

-Peff