Re: state of the 2.6.13 transition

On Tuesday 20 September 2005 04:08 pm, Phil Blundell wrote:
> On Tue, 2005-09-20 at 21:55 +0100, Ian Molton wrote:
> > I've patched a lot of stuff but havent comitted it because I cant test
> > it. it seems that HEAD is "missing" some files needed to build and I
> > dont understand why.
> >
> > If anyone could explain it, I'd love to know the reason...
> >
> >From what I gather, the reason that the HEAD is broken is that 2.6.13
>
> has been imported onto the vendor branch but not yet fully merged into
> the trunk.  This is just the way that vendor branches in CVS work: it is
> a bit unfortunate, but you have to live with it.  Essentially, between
> the time that you do "cvs import" and the time that you commit the final
> merge result, the tree will be in an inconsistent state and most likely
> broken.  To minimize disruption, you need to either (a) keep that time
> as short as possible, or (b) do the merge entirely in a working tree,
> without importing onto a vendor branch at all, in which case the HEAD is
> not affected at all until you commit the changes.

One more detail: I think the breakage relates to files that exist only on the 
vendor branch; i.e., files that we haven't modified from upstream.

For example, take arch/arm/lib/gcclib.h. The old HEAD version of that file has 
never been modified, so the HEAD version that gets pulled is the latest on 
the vendor branch. Now that 2.6.13 has been imported, the latest version of 
that file from the vendor branch is the 2.6.13 version, which in this case 
breaks the build when most of the rest of the sources are from 2.6.12.

Matt