Re: [PATCH 2/2] transport-helper: update remote helper namespace

From: Felipe Contreras <felipe.contreras@gmail.com>
Date: Apr 13, 2013, 06:00 UTC
Message-ID: <CAMP44s3+eaRbeXP0bPXMSE8Z1K_Lqyu8e1XCvudkapkTJFBWnA@mail.gmail.com>
In-Reply-To: <CAMP44s1LF46VU0E4W=r-qog3JY+Y-qyYGfkqxLEnTG8X8GoWOA@mail.gmail.com>

Hi,

Why wasn't this patch merged to 'pu'? To my knowledge nobody raised any real concerns.

Should I explain in every commit that touches transport-helper how remote-helpers without marks are impossible? I know I said I was going to update the commit message, but I don't think that reason to not put it in 'pu'. Also, the only reason I said so was to make Jeff happy, but now that I think again, it doesn't really belong there; remote helpers cannot be using these refs, they just can't. They cannot work without marks, it's not possible. To think otherwise is simply a mistake.

On Thu, Apr 11, 2013 at 12:18 AM, Felipe Contreras <felipe.contreras@gmail.com> wrote:

Show 38 quoted lines

> On Thu, Apr 11, 2013 at 12:05 AM, Jeff King <peff@peff.net> wrote:
>> On Wed, Apr 10, 2013 at 11:53:38PM -0500, Felipe Contreras wrote:
>>
>>> > But if we push some commits to the helper, moving Y up to Z, then it
>>> > would build the new commit (which contains the foreign-vcs's equivalent of
>>> > Y..Z) on top of Z, not Y.
>>>
>>> Why would it do that? If X points to say revision 100, presumably it
>>> was stored somewhere while doing a fetch. Similarly, if foreign
>>> version of Z is 150, it can update that number while doing a push. The
>>> next fetch it would start from 151.
>>
>> I think the only reason not to bump the marker forward during the push
>> would be if the helper wants for some reason to "re-import" from the
>> foreign source rather than accepting the git versions of the commits.
>> Something like git-svn's markup of the commit messages with revision ids
>> comes to mind.
>
> Yeah, but that's already a second level hypothesis. First,
> remote-helpers would need to be able to work without marks, and they
> can't.
>
>> But if it matters, then by definition that would mean
>> that the import/export is not bidirectionally clean.
>
> I don't see how would that matter.
>
>> So I can buy the argument that bumping it forward ourselves will not
>> matter for any well-implemented helper.
>
> Or any helper.
>
>> That is the sort of thing that might be helpful to include in the commit
>> message; if somebody does run across such a helper and bisects to your
>> commit, then they can understand the rationale for the decision.
>
> If it did matter, it would be mentioned. I will updated it later if
> there's no further comments.

-- 
Felipe Contreras