[ewg] Re: [ofa-general] Fwd: Re: using stgit/guilt for public branches

[Michael S. Tsirkin]

> Quoting Sean Hefty <sean.hefty at intel.com>:
> Subject: RE: [ofa-general] Fwd: Re: using stgit/guilt for public branches
> 
> >FYI. I posted a question on git mailing list, asking about
> >best ways to manage ofed repository.
> >
> >http://article.gmane.org/gmane.comp.version-control.git/45519
> >
> >The conclusion so far seems to be that what we are doing (keeping patches
> >under version control) is basically the right way to do it:
> >
> >http://article.gmane.org/gmane.comp.version-control.git/45569
> 
> I strongly prefer that the patches be applied to the actual code.

It looks nice on the surface, but won't work well in practice I'm afraid.
There's only one way to get close to this - agree that OFED, as a rule, should
not include code that is not upstream on kernel.org 
(we could add out of kernel modules without touching core, but that would be all).
I would be fine with this rule, but would you?

> Someone who wants to use stgit can generate their own set of patches and
> manage them off the tree if they want.

This managing of two parallel trees would fall on the shoulders
of OFED maintainers, and I don't think it's practical to keep it up
in parallel with OFED integration which is also a full time job.

> Right now it's way too difficult to see what code is there,

How is it difficult?
"What code is there" is not well-defined, since it actually
depends on the distro.
For a specific distro - get the tarball, run ./configure, and look.

> switch 'branches',

What does this mean?
Do you want to branch off ofed 1.2?
git-branch and off you go - build scripts *already*
can get a branch name.
Again, what's so difficult?

> generate patches,

Did you read the howto's?
it's *really* easy to do using quilt, I do it all the time:
quilt new foo.patch
quilt add <file>
quilt refresh

> etc.

I have to ask - did you read the actual thread?
You seem to ignore all arguments why using stg won't work for a public tree.
In this thread, git guys told us keeping patches under git as we already do
seems to be the best option available option.

> We're doing everything manually and losing the majority
> of the benefits that git gives us.

This is clearly not true. We are pulling code from multiple people
completely automatically by git pull. We have a build system that
can be pointed at any git tree and will generate a tarball from
there. There's a hash based checksum that identifies each build
in a unique way. All these are main benefits of git.
And while git does not include tools to apply patches for you,
patches are also applied by quilt, not manually.

So - 
What is done manually? What benefits did we lose?

-- 
MST