On Tue, Jun 16, 2015 at 1:54 AM, Junio C Hamano <gitster@xxxxxxxxx> wrote:
> Paul Tan <pyokagan@xxxxxxxxx> writes:
>
>>> The scripted Porcelain is spawned after applying patches 1-3 from
>>> here, when you do not have _GIT_USE_BUILTIN_AM exported. Haven't
>>> RUN_SETUP code did its thing by that time?
>>
>> Ah right, the RUN_SETUP code would have chdir()-ed to the working
>> directory root, so git-am.sh will be unable to find the original
>> working directory. To aid it, we would have to chdir() back to the
>> original working directory, and unset GIT_DIR.
>
> I do not think that is a correct workaround, though. GIT_DIR may
> have come from the end user, i.e.
>
> $ GIT_WORK_TREE=somewhere GIT_DIR=somewhere.else git am ...
Ah, forgot about that ><
> As the RUN_SETUP|REQUIRE_WORK_TREE bit is merely a convenence in
> git.c, one workable way to keep these dual implementations is to do
> what built-in commands used to do before these were invented.
> Perhaps studying how cmd_diff(), which is run from git.c without
> either RUN_SETUP or NEED_WORK_TREE, and taking good bits from it
> would help. I think the implementation roughly would look like
> this:
>
> int cmd_am(int ac, const char **av, const char *prefix)
> {
> /*
> * NEEDSWORK: once we retire the dual-mode
> * implementation, this preamble can be removed...
> */
> if (... want to do scripted ...) {
> ... spawn the scripted thing ...
> }
> prefix = setup_git_directory();
> setup_work_tree();
> /* ... up to this point */
>
> ... your real "git am in C" comes here ...
> }
>
Ah OK. Just to be sure, I took a look at run_builtin() in git.c, and
indeed it only just does:
prefix = setup_git_directory();
trace_repo_setup(prefix);
setup_work_tree();
which is not bad at all. Thanks.
Regards,
Paul
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html