[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Quilt-dev] Using Quilt with other SCM
From: |
Josh Boyer |
Subject: |
Re: [Quilt-dev] Using Quilt with other SCM |
Date: |
Sun, 26 Feb 2006 17:24:24 -0500 |
On Sun, 2006-02-26 at 09:47 +0100, Jean Delvare wrote:
> Hi Thomas, Josh,
>
> > > I'm using Quilt to prepare a stack of patches that I then selectively
> > > commit (when there are ready) to the non-distributed SCM of my team. Of
> > > course, I only commit the patches that are at the bottom of the stack:
> > > (...)
> > > Would it be possible to have a command like "quilt drop" (or an option
> > > to "quilt delete") that removes a patch from the series without actually
> > > poping it and touching the working copy ?
> >
> > I sent a patch for something like this quite a while ago. There was
> > some decent discussion on it, but I haven't had time to really address
> > some of the issues that were brought up.
> >
> > Perhaps if you look back in the archives you might find something worth
> > while.
>
> If I remember properly, the main issue with your proposed
> implementation was that it wasn't user-error-proof enough. It was
> assuming that the user was doing exactly the right thing, rather than
> properly preventing him/her from screwing up.
Yep, that is what I remember too.
> This drop command can only be added to quilt if it is safe to use. This
> suggests that the following two assertions are verified:
> 1* Only the bottom-most patch can be dropped.
> 2* No other applied patch further in the series must modify any file
> modified by the to-be-dropped patch.
Yep, I still agree with that.
> Rule 1 could be extended to the N bottom-most patches (if the user wants
> to merge several quilt patches and commit them all at once in his/her
> SCM of choice). However, it is probably safer to only let the user drop
> one patch at once, and have him/her run the command as many times as
> needed.
It would also make the implementation simpler as well, which is a good
thing.
> Rule 2 may be easily enforced by only allowing the drop command to be
> used when no other patches are applied.
Right. So stacks of patches would have to be "dropped" in order. I
think that is a good thing.
> We also want an option to backup the dropped patch, like the delete
> command has.
Oh, yeah good idea. I didn't think about that one.
Like I said earlier, it's mostly a lack of time that has prevented me
from cleaning things up. That, and my bash skills aren't nearly as
complete as my other programming skills. I'll try and find some time
soon, but no guarantees :)
josh