[issue99] mq: support mq repositories

[Samuel Masham]

On 03/01/06, Vadim Gelfer <vadim.gelfer at gmail.com> wrote:
> On 1/2/06, Samuel Masham <samuel.masham at gmail.com> wrote:
>
> > > it would be very good if mq supported repositories properly. i mean that you
> > > should be able to revision-control mq patch sets and series files using hg.
> >
> > You can its just by hand at the moment.
>
> i know but that makes a lot of extra typing.

:)

Also I think we should be able to do this from a simple wrapper script*

> > Well I have also done some work on this and have a patch that will
> > version the patch repos automatically if configured to do so.
>
> what does this mean?

as you work on the patches and the queue all changes are versioned in
the background.

If you are publishing the patch queue you may want to use this as a
working repos and have a separate public one allowing you to prune the
uninteresting versions.

The idea is that the patch queue is visioned without any interaction
from the user.

> > Also instead of adding a ci command adding a -c option to existing
> > commands was favored by Chris
>
> if you mean that command like qrefresh should have -c option, i do not
> agree. let me give example.
>
> i work on patch series against kernel 2.6.13. get it into ok shape.
> check in all patches at one time with qci (= "commit changes in patch
> repo"). pop all patches. pull kernel 2.6.14 into main repo. push
> everything, fix rejects in each patch as i go.
>
> i want to check in new set of patches at end of all this work as
> atomic change, not one by one as i fix every one.

... and a single qrefresh -c at then end would do this.

The issue is that with the mq commands there is a danger that you just
mirror ALL of the hg commands but working on the patch queue.

(The Chris I mentioned is Chris Mason and he wants to avoid this duplication)

As for if the qci command, is is an obvious candidate but is there
really any need for this to be separate?

> > Finally in the mq repos it is sometimes useful to version the status
> > file etc so hard coding the ignore file doesn't fit with my usage at
> > least.
>
> it is trivial to fix my patch to not touch ignore file if is already
> existing. no problem.

I thought your version created the repos with the ignore file...

Samuel

(*still no access to hg so can't run or try anything ...)