Re: [Caml-list] Trivial compiler patches

[Anil Madhavapeddy <anil@recoil.org>]

On 25 Mar 2014, at 16:16, Richard W.M. Jones <rich@annexia.org> wrote:

> I have to say that a number of projects I've been involved with have
> rejected github pulls as a method of working on patches.  I think the
> main reasons are:
> 
> - A mailing list allows you to use your regular editor when
>   commenting.
> 
> - A mailing list provides a text-based, open archive of historical
>   discussions of patches.

These two are definitely useful, although it substantially increases
the amount of incoming email for all developers (especially with 
large, repeated patchsets that are being iterated over).

> - github pulls always(?) turn into merge commits, instead of a
>   linear history.

Only if they're merged via the web interface.  They can be rebased
and pushed directly as usual from the command line.

> - github is closed source

Certainly a concern.

> 
> - AFAIK github is not integrated into any CI system (compare, say,
>   Gerrit + Jenkins -- although Gerrit + Jenkins have their share of
>   problems as well).

GitHub has fantastic CI support via Travis -- easily the best of
anything I've used in the past, and particularly so because it's
free and hosted elsewhere.

See: http://anil.recoil.org/2013/09/30/travis-and-ocaml.html
for an overview of getting started with it.

In general, the GitHub API support makes it sufficiently easy to
hook in third-party workflows that the OCaml GitHub mirror will
continue to be useful even if Gabriel's experiment with pull requests
doesn't work out.

> Personally I don't think that patch review is anywhere close to being
> a solved problem.  It's an important area requiring much more
> research!

I spent a while trying these out for Real World OCaml, and noted
that all of the existing tools only support *patch* review, but not
general *code* review.  Consider, for example, someone reading a
book chapter, or the 4.01.0 x86_64 code generator and wanting to leave
comments on the whole tree, not just on a specific isolated patch.

An odd omission in the space, or I just missed an area of tooling
that deals with this...

-anil