llvm dev - Jul 2016 - [RFC] One or many git repositories?

> On Jul 26, 2016, at 10:32 AM, Renato Golin <renato.golin at
linaro.org> wrote:
> 
> On 26 July 2016 at 17:59, Mehdi Amini <mehdi.amini at apple.com>
wrote:
>> When exporting a commit from the monorepo to one of the read-only
>> repository, it should be possible to generate a “git-svn-id”-like the
same
>> way we’re doing when exporting from SVN today.
> 
> Ha! Yes it does! Kind of reverse sub-modules. Now I get it.
> 
> However, this would require some scripting, right?
Yes, just like we do when exporting from SVN today I think. 
(The script might be slightly less complex since git is natively handling the
export we’re doing today).
> 
> AFAIK, GitHub doesn't support hooks at that level, so we'd either
have
> to host it somewhere else or use some more convoluted solution, no?
Since it is a deterministic process (even if the update does not happen for an
hour, it can be replayed) in this direction (monorepo -> multiple subrepo
contrary to multiple repo -> monorepo) it can be done asynchronously (i.e.
using a cron or a web hook).

— 
Mehdi

On 26 July 2016 at 18:36, Mehdi Amini <mehdi.amini at apple.com>
wrote:
> Yes, just like we do when exporting from SVN today I think.
> (The script might be slightly less complex since git is natively handling
the export we’re doing today).
And this would be on a similar level of complexity than sub-modules
update (just the script), I assume. Some hook/webhook that can delay
because the "source-of-truth" is somewhere else.

The better thing in this scenario is that the "source-of-truth" is
also the "well organised one", whereas in the sub-modules it's the
other way round.

cheers,
--renato

> On Jul 26, 2016, at 10:59 AM, Renato Golin <renato.golin at
linaro.org> wrote:
> 
> On 26 July 2016 at 18:36, Mehdi Amini <mehdi.amini at apple.com>
wrote:
>> Yes, just like we do when exporting from SVN today I think.
>> (The script might be slightly less complex since git is natively
handling the export we’re doing today).
> 
> And this would be on a similar level of complexity than sub-modules
> update (just the script), I assume. Some hook/webhook that can delay
> because the "source-of-truth" is somewhere else.
> 
> The better thing in this scenario is that the "source-of-truth"
is
> also the "well organised one", whereas in the sub-modules
it's the
> other way round.
You still missed the *major* point: the process can be asynchronous and be
deterministic.

— 
Mehdi