Labels: Type-Enhancement Priority-Medium
New issue 2350 by stephenm...@gmail.com: Need support for stgit and/or quilt
*NOTE: If you have a patch, please submit it to
What version are you running?
What's the URL of the page this enhancement relates to, if any?
Describe the enhancement and the motivation for it.
Stacked git is a patch management system built on top of git.
quilt is another patch management system (inferior to stgit, last I
We're also using subversion (svn) as our main repository.
So far, the way I work has been like so:
Checkout working directory from svn.
Create a new git repository in my working directory and check everything in
local git repository.
Put stgit on top of that. So that's stgit on top of git on top of an svn
hack, hack, hack away and make a bunch of patches. (The patches are stored
as git commits by stgit in the local git repository.)
As I get the patches to my liking, I can commit them back to svn.
So far so good, stgit allows me to "polish" the patches so that when
someone digs through the svn history, I look like someone who seems to know
what they were doing, with not too much backtracking, partial working
commits, etc. (Backing up intermediate work is done via stg export.)
Now I has a problem.
I might have 20 or so patches at any given time (which is not really a
large number -- a large number would be in the 1000s.)
For example, currently I have these patches:
[scameron@localhost linux-3.1]$ stg series
Many are independent of one another, and these are no problem. I can
generate an svn diff which reviewboard is happy with by rearranging the
order of the independent patches with stg such that each patch is in turn
moved to the first position in the stack, then "svn diff" gives me a good
diff. It's a pain in the ass, but workable.
The problem comes when I have patches that depend on one another.
Reviewboard does seem to have the concept of a "parent diff", and you can
get to this (by a bizarre seeming sequence of moves in which you create a
review, give it some diff it will be happy with, then edit the review to
update the diff, and when you update the diff, it gives you the chance to
specify a "parent diff".) However, reviewboard doesn't like the format of
diffs from "stg show" or git diff, for some reason. (If it did, my problem
would be solved, I think, as I could use "stg show" to get my dependent
diff, and "git diff" to get the combined diffs which it depends on
(the "parent diff") and feed those to review board.)
If I have a patch X, which depends on patch A, B, and C, then I think I can
get the "parent diff" by pushing on A, B, and C, then doing "svn diff". I
can't get the patch X itself in a usable form though. "stg show" will get
it, and I can also get it with "git show (with the commit hash)" but these
are in a form which reviewboard seems unable to comprehend. Svn can't get
this diff, except conglomerated together with the "parent diff".
Consequently, Reviewboard ends up serializing the review process at best,
as I then have to commit the "parent diff" to svn in order to even review
the diffs that depend on the parent diff. That's not really workable for me.
What operating system are you using? What browser?
Linux, various flavors, firefox various versions.
Please provide any additional information below.
You received this message because you are subscribed to the Google Groups
To post to this group, send email to email@example.com.
To unsubscribe from this group, send email to
For more options, visit this group at