[Gnu-arch-users] Re: integration still broken

[Aaron Bentley]

On Sun, 2004-01-18 at 07:20, Robert Collins wrote:
> On Fri, 2004-01-16 at 10:31, Tom Lord wrote:
> > No can do, pardner.  You're branch commits the moral sin of BREAKING
> > MKPATCH.
> 
> Right. We've got it down to two separate breakage:
> 
> 1) inode-sigs are -already- broken.
> I've added a test case to pick this up if any patch logic depends on
> inode sigs in the future, and disabled the perf tweak that used it.
> 2) mkpatch is differently fucked.
> I've setup the changeset-tests to run in-build-tree automatically.
> 
> abently is on a fix to 1's root cause, AIUI. I'm tracking down 2).

Actually, I'm not working on root causes.  I've worked around the broken
inode-sigs.  In my tree, the only case that isn't handled properly is
the case where a change is not reflected in the mtime or size, but
that's not a new problem-- it's in Tom's tree too, and that problem
cannot be solved without killing the performance enhancement of
inode-sigs.

On the plus side, those who get --link won't notice the loss, since it's
easy to determine that two hardlinks to the same inode have the same
contents.

What I'm working on right now is trying to sync my tree with
barch/tla--integration.  I still don't have patches 51 through 60.

Aaron