Re: [Gnu-arch-users] Re: [PATCH] Re: bazaar bug #385 (Re: Stabilizing Bazaar?)

[Robert Collins]

On Sun, 2005-05-22 at 23:13 +0200, Matthieu Moy wrote:
> Robert Widhopf-Fenk <address@hidden> writes:
> 
> > I did
> >
> >   baz register-archive http://bazaar.canonical.com/archives/address@hidden
> >   baz get address@hidden/bazaar--pristine-version--1.4--patch-2
> >
> > but I am not sure how to built from your tree, so I did
> >
> >   baz get 
> > http://bazaar.canonical.com/archives/address@hidden/dists--bazaar--1.4 
> > bazaar
> >
> > and replaced bazaar/src/baz by your tree and built by
> 
> Alternatively, you could have used "baz replay" to cherrypick the
> patch from me (In Xtla: C-u M-x baz-missing RET RET
> address@hidden/bazaar--pristine-version--1.4 RET, mark
> with 'm' and replay with 'r'), or "baz switch" (M-x baz-switch RET) as
> an improved shortcut for "rm + baz get".
> 
> I've continued the idea behind my patch (preset version and revision
> fields for project tree objects representing pristine).
> 
>   address@hidden/bazaar--pristine-version-submit--1.4--patch-2
> 
> That I've submitted to the pqm, so it should appear soon in the
> official archive.

I'm afraid I've raced with you on this one - I spent the weekend largely
offline destressing and doing preparatory work for the win32 revlib
formats.

Anyway, I've added a check to arch_find_local_tree_copy (which all
reference tree lookups go though) to validate the tree metadata. If the
tree does not match what we expected (both revision and default-version)
then its removed and a new one created.

I considered just setting it in memory, or just overriding it, but we
really can't guess why it will be wrong - there are known causes, but it
could be something more serious - so this is the safest route.

Cheers,
Rob

signature.asc
Description: This is a digitally signed message part