[llvmlinux] Using quilt instead of python scripts

Behan Webster behanw at converseincode.com
Wed Sep 12 22:46:00 UTC 2012

On Wed, Sep 12, 2012 at 5:05 PM, Mark Charlebois <charlebm at gmail.com> wrote:

> The recent change to the build system to use quilt has broken msm and
> would likely break any non-tip Linux build (i.e. raspberry pi).
That was an unintended side effect. Mea culpa.

> The old system used to allow upstream patches that didn't apply to be
> filtered at the patch hunk level. Currently if a patch hunk does not apply
> then the patching process fails and there is no easy way forward.
It will only really work when we get all the builds working from HEAD. But
that won't happen anytime soon. :(

One advantage to using something like quilt is that the patches are
designed to apply cleanly. One of the issues we had before was that people
couldn't necessarily take our patches and use them elsewhere without the
same kind of scripts to remove parts of the patches. That kind of makes it
difficult to upstream clean patches...

> I am all for using quilt as long as it doesn't remove the previous
> functionality.
My intention wasn't to remove functionality, though I have to admit I added
it rather more quickly than I should have. Again, my bad.

I think we needed to add the ability to support patches for other kernel
trees. Or indeed, the arch/all patches maybe needs to be per kernel tree.
Afterall, if the code in a kernel tree diverges from mainline, or is old
enough, merely removing some of a patch if it doesn't apply doesn't seem
like the appropriate action to take. Rather better to have a patch which is
meant for that tree which does the same thing as the patch which failed.

I think we can make it work.

Behan Webster
behanw at converseincode.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/llvmlinux/attachments/20120912/51ac4da0/attachment-0001.html>

More information about the LLVMLinux mailing list