RFC: Package Review VCS

Josh Boyer jwboyer at gmail.com
Tue Nov 20 11:59:58 UTC 2007


On Mon, 19 Nov 2007 22:39:08 -0500
Warren Togami <wtogami at redhat.com> wrote:

> Josh Boyer wrote:
> > 
> > The idea behind this is nice.  Using CVS to do it will suck.  
> 
> What about it exactly?
> Using CVS as a tool again when we should be using a more modern VCS?

Yes.  The fact that you manually have to copy the history, which isn't
all that important anyway, from one repo to another sucks and adds one
more barrier before the package is in.

> > Also, I'm not sure it buys you a ton over keeping track of changes in the
> > specfile during review, which should already be done anyway.
> > 
> > josh
> > 
> 
> While most spec changes during a review will be boring, it is sometimes 
> worthwhile to keep them in history.  A more interesting thing this 
> enables may be during pre-review, where it can act as a playpen/staging 
> ground for rapid packaging and testing.  Imagine this:
> 
> 1) Existing Fedora packager begins work on a new package.
> 2) cvs-import.sh in /cvs/pkgreview, existin Fedora packager can create 
> whatever they want, whenever they want.
> 3) They might ask other people to help in packaging.  *Real* quick 
> centralized place to point other people to ask for help.

Help how?  One of the reason for reviews is to see how responsive and
adaptive the packager is.  Others can make suggestions, but shouldn't
be making changes.  That should rest squarely on the packagers
shoulders alone.

> 4) *Real* quick build testing on all archs directly from your 
> work-in-progress.
> 
> I call that enabling.
> 
> I suppose we could make using /cvs/pkgreview optional for package 
> reviews.  And Merge Review could use /cvs/pkgs instead.  But I can't 
> imagine why anybody would prefer the hassle of spinning new .src.rpm's 
> and uploading them to a web accessible URL several times instead of 
> simply using a VCS.

I don't particularly care about whether it exists or not.  I just think
that:

1) Changes during review should be reflected in the spec file, and
2) That history isn't all that important and doesn't need to be
manually copied in repo files to the official package VCS.

josh




More information about the devel mailing list