Re: hggroup available in graphics archive

[Thomas Weber]

Am Mittwoch, den 30.04.2008, 10:37 +0200 schrieb Michael Goffioul:
> On Wed, Apr 30, 2008 at 9:17 AM, Thomas Weber
> <address@hidden> wrote:
> >  > Recently, I tried to use MQ locally. This revealed very handy to work
> >  > on a single patch incrementally, even on multiple development
> >  > setup (I use 2 computers for development): I just had to copy the
> >  > patch between the 2 computers and play with qpop/qpush. In the
> >  > end, I commited a larger, but more atomic, changeset to the
> >  > archive. However, I'm not sure how to deal with that in the context
> >  > of a shared public archive.
> >
> >  I'd say you shouldn't. Keep your patches small, so individual changes
> >  can be pulled. But don't throw away most of the machinery your VCS
> >  offers you, especially when collaborating with others.
> 
> If I understand correctly, going with the MQ way while still keeping the
> possibility to share an archive and make it public would mean the
> following:
> 1) the graphics archive would be a clone (at some point in time) of the
> main octave archive, along with a set of patches that are under
> revision (the patches are unapplied to the graphics archive)
> 2) we never commit the patches ourself to the archive (hg qdelete -r), but
> we wait for the patch inclusion in the main archive, then synchronize the
> graphics archive, then "hg qdelete"
> 3) people wanting to access the graphics code would have to grab the
> graphics archive and "hg qpush -a" to apply the patches, then compile
> 
> Am I right?

Yes. 

I don't think you gain much that way. If a qpatch is not fine-grained
enough, you have to split it by hand. That's about the same as pulling
the same changes as a normal patch and removing the stuff you don't want
by hand and commit a second patch afterwards.

        Thomas