Re: [PATCHv1] completion: add option '--recurse-submodules' to 'git clone'

[Stefan Beller <sbeller@google.com>]

On Thu, Jul 28, 2016 at 9:22 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Stefan Beller <sbeller@google.com> writes:
>
>>> Anyway, I'll apply the "addition to the completion" patch.
>>>
>>> Thanks.
>>
>> Thanks for this patch!
>>
>> Note: if we ever decide to resurrect sb/submodule-default-path,
>> we run into a merge conflict. The reasoning for using
>> "--recurse-submodules" instead of a plain "--recurse" makes sense
>> as well, so that merge conflict will be resolved in favor of this patch.
>
> Thanks for an advance warning.  My rerere database has already been
> taught about this conflict ;-)
>
> As to sb/submodule-default-path topic, which has been blocked on
> still-in-flux attribute work, I am tempted to declare that the
> attribute work is not yet thread-ready but it is in a good enough
> shape to base other works on, and have them advance to 'next'.

I am tempted to ask for delaying sb/submodule-default-path
a bit more and see if there is an alternative solution.

Inspired by Duys series on submodules and worktree I found a new
way how to do the submodule management (separation of the submodule
URL as a flag indicating the existence and the actual URL), and that
will be very similar to what is implemented in sb/submodule-default-path
just even more streamlined. (That said the state of that new series
is still vapor and not yet solid, so ... it's hopes and dreams.)

>
> The traditional pattern of allowing the callers to randomly allocate
> an array of "struct git_attr_check" and passing the pointer to its
> first element to git_check_attr() function was impossible to extend
> without having to update the callers, but we have migrated away from
> the pattern and the attribute subsystem can be enhanced without
> impacting the callers too much.
>