Re: [PATCH v2 00/46] "struct pathspec" conversion and :(glob) and :(icase)

From: Junio C Hamano <gitster@pobox.com>
Date: Jul 15, 2013, 21:43 UTC
Message-ID: <7va9lna48n.fsf@alter.siamese.dyndns.org>
In-Reply-To: <1373790969-13000-1-git-send-email-pclouds@gmail.com>

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

Show 11 quoted lines

> Compared to the last round [1] this series mainly fixes comments and
> commit messages suggested by Eric and Junio. It also fixes a conflict
> with cb/log-follow-with-combined (in master) and introduces :(icase)
> mentioned in the last round.
>
> [1] http://thread.gmane.org/gmane.comp.version-control.git/226892
>
> Nguyễn Thái Ngọc Duy (46):
>   clean: remove unused variable "seen"
>   ...
>   parse_pathspec: accept :(icase)path syntax

This was a good exercise for git-imerge. I merged it just below the "2.0 migration topic" that is kept outside the 'next' first by hand, and then used imerge. Interested parties may try to do this, after fetching from me:

	$ git checkout cb5d2fc
	$ git branch nd/magic-pathspec 93d93537
	$ git imerge start --first-parent --name bogo nd/magic-pathspec

	... imerge thinks and then will stop, asking you to resolve
	... and commit, repeatedly.  Do the next two steps until it
        ... stops.

	$ edit
        $ git commit -a --no-edit && git imerge continue

	... and then
        $ git imerge finish

The imerge program was a bit under-help (yet), and also it was a bit cumbersome at times having to resolve trivial and obvious two conflicts in the same file (e.g. the series adds description for GIT_*_PATHSPECS environment variables to Documentation/git.txt in two separate patches, and the resolution is obvious for a single merge, but imerge wants me to resolve them in two steps), but in the end, after "imerge finish", it produced identical result as my hand merge, so at least it seems to be doing something right ;-)

A few things I noticed:

 - It was not immediately clear from "git imerge -h" text what the
   simplest "hello world" workflow was.

 - It was not clear what it wanted for --name parameter (it becomes
   the name of the temporary and final branch that stores the
   result).

 - The final step "imerge finish" gave me this ugliness:

     Merge commit 93d9353... into commit cb5d2fc7

   Perhaps you can at least use the initial branch name
   "nd/magic-pathspec" I gave you, and use "git fmt-merge-msg"?

A bigger test will be to try merging jn/add-2.0-u-A-sans-pathspec and jc/add-2.0-ignore-removal topics on top of the merge result (2857023 in today's pushout). These two merges, especially the latter one, have to involve some "evil merges", so it would be interesting to see how well it fares.

In any case, thanks for a fun tool.