[dmd-internals] To consume pull requests for trivial bugs

Leandro Lucarella luca at llucax.com.ar
Tue Aug 21 06:20:32 PDT 2012

Andrei Alexandrescu, el 20 de August a las 14:39 me escribiste:
> On 8/20/12 2:31 PM, David Nadlinger wrote:
> >On Mon, Aug 20, 2012 at 8:09 PM, Brad Roberts<braddr at puremagic.com>  wrote:
> >>Sorry, but the fix to not enough people with front end experience doing
> >>code reviews is not to not do code reviews.
> >
> >Agreed, but I'd like to add that this is a general problem with DMD
> >development (maintainability?) - even Walter regularly commits code
> >which doesn't even pass the auto tester.
> This issue is a symptom of a simple underlying problem: we don't
> have enough contributors. Kenji tried to find a solution within the
> current constraints, whereas Brad responded from an "enough
> reviewers" frame of reference.
> I think we need to add more active committers to dmd.

That's not really necessary, all you need is Walter trusting other
people and then *pulling* from other people's repos. That way he can
even verify the changes before *pushing* to the public repo. The
workflow for Walter can be:

1) pull from trusted contributor
2) verify (run the tests, locally review the changes, maybe just read
   the commit title to see what kind of changes are being pulled, or
   whatever, including doing nothing)
3) pushing to the public repo (or not, if 2) have problems)

This greatly reduces Walter's job, but he can still have full control on
*when* things get pushed to the public repo. For example, when
a stabilization process starts, is not so important to ask the
contributors to stop pushing big changes, everyone can keep doing
whatever they want without messing with the public repo.

I think the main problem here is Walter having to review and test every
single commit. If he trust contributor, he can pull from then and
introduce a batch of commits with just a minimal verification process
for the whole batch, instead of having to go through them one by one.
For me having a lot of people pushing to a repo is too risky, to easy to
make a mess of it (BTW this is how Linux development process works, only
Linus pushes to the public vanilla kernel repo, but he pulls from people
he trust, which in turn pull from the people they trust and that's how
the model scales up with a "tree of trust").

Leandro Lucarella (AKA luca)                     http://llucax.com.ar/
GPG Key: 5F5A8D05 (F8CD F9A7 BF00 5431 4145  104C 949E BFB6 5F5A 8D05)
Y tuve amores, que fue uno sólo
El que me dejó de a pie y me enseñó todo...Bien

More information about the dmd-internals mailing list