Re: [PATCH v15 7/7] t/t7507: tests for broken behavior of status

On Tue, May 3, 2016 at 5:18 AM, Pranit Bauva [off-list ref] wrote:

On Tue, May 3, 2016 at 12:19 PM, Eric Sunshine [off-list ref] wrote:

quoted

quoted

quoted

Step back a moment and recall why these tests were added. Earlier
rounds of this series were buggy and caused regressions in git-status.
As a consequence, reviewers suggested[1,2] that you improve test
coverage to ensure that such breakage is caught early.

The point of these new tests is to prevent regressions caused by
*subsequent* changes, which is why it was suggested that these tests
be added early (as a "preparatory patch"[3]), not at the very end of
the series as done here in v15.

Sure! I just wanted the commit message to be detailed as per the
guidelines given by SubmittingPatches. I will swap the patch 6/7 and
patch 7/7 changing the commit message. Also I will make the commit
message less detailed.

This patch should be inserted before 4/7 since it needs to protect
against breakage which might occur when 4/7 changes the behavior of
OPTION_COUNTUP.

I forgot to mention about this earlier. When I was rebasing, this stroked me.
I guess making any changes in ordering the commits will make one of
the test as absurd. One of the test uses a configuration variable
'commit.verbose' will won't be effective before the patch 6/7. So I
guess I will have to only change the commit message to reflect as
"improving test coverage".

I also had intended to talk about this but forgot. What would be quite
logical is to introduce only the "git-status without --verbose" test
in this new "improve coverage" patch before 4/7. The other test, which
ensures that git-status doesn't regress with commit.verbose, would
then very naturally be included in the patch which adds the
commit.verbose functionality (currently patch 6/7).