On Thu, Aug 18, 2016 at 11:46 AM, Junio C Hamano <gits...@pobox.com> wrote:
> Jacob Keller <jacob.e.kel...@intel.com> writes:
>> From: Jacob Keller <jacob.kel...@gmail.com>
>> Currently, do_submodule_path relies on read_gitfile, which will die() if
>> it can't read from the specified gitfile. Unfortunately, this means that
>> do_submodule_path will not work when given the path to a submodule which
>> is checked out directly, such as a newly added submodule which you
>> cloned and then "git submodule add".
> Hmm, are you sure about that?
> A call to read_gitfile() turns into a call to read_gitfile_gently()
> with the return_error_code parameter set to NULL. The function does
> a stat(2), and if the given path is not a file (e.g. we created the
> submodule working tree and repository in-place ourselves, instead of
> cloning an existing project from elsewhere, in which case we would
> see a directory there), it says READ_GIT_FILE_ERR_NOT_A_FILE and
> returns NULL, because that is not a fatal error condition. The same
> thing happens if path does not yet exist.
> This caller is given <path>, prepares "<path>/.git" in buf, and
> calls read_gitfile(). If it returns a non-NULL, it replaces what is
> in buf and continues, but if it returns a NULL (i.e. the two cases I
> mentioned in the above paragraph), it continues with "<path>/.git".
> While I do not think changing it to resolve_gitdir() is wrong per-se,
> I am not sure if it is necessary.
> I must be misreading something in the existing code.
No I think you're correct. I was under the assumption that it would
die here. I think it's probably better to stick with read_gitfile()
here, it should work. The main issue is what happens after this needs
to be fixed.
I'll investigate this more.
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html