Re: [Gnu-arch-users] Re: conflict applying patch in arch_build_revision

[Jan Hudec]

On Tue, Nov 09, 2004 at 11:11:24 +0100, Matthieu Moy wrote:
> Masatake YAMATO <address@hidden> writes:
> 
> >>     address@hidden jet]$ tla get address@hidden/gtracer--mainline--0.0 
> >> /tmp/examples
> >>     * from pristine cache: address@hidden/gtracer--mainline--0.0--base-0
> >>     * patching for revision: address@hidden/gtracer--mainline--0.0--patch-1
> >>     PANIC: conflict applying patch in arch_build_revision
> 
> Seems your archive is corrupted :-(

It does not have to be the archive. It might be the "pristine cache",
that is pristine tree in your working copy. Try to remove each and every
{arch}/++pristine-trees directory you can find. In fact, those in
/tmp/* should suffice.

By the way, is the machine running tmpreaper? That would explain
corruption of pristines under tmp.

> Normally, the changeset patch-N+1 should apply exactly to the revision
> patch-N, because it was uploaded by "tla commit", which computed the
> changeset based on patch-N (which was in a pristine tree or in your
> revlib at that moment).
> 
> If you had a corrupt revlib or library when you comitted (this
> happened to a friend once because we had changed history : Lost a few
> changesets after an unfortunate "rm -rf ~" -- don't laugh -- and
> rebuilt the changesets without unregistering the revlib), then, this
> is no longer the case, and the inexact patching may result in
> conflicts, and thus, PANIC.
> 
> Solutions :
> 
> * Manually edit the changesets and upload the modified version to your
>   archive (VERY dangerous),
> 
> * Start a new branch and never use this one again,
> 
> * Is there something better?

No, there is not. But the panic does not necessarily mean things are
this bad and they likely are not!

-------------------------------------------------------------------------------
                                                 Jan 'Bulb' Hudec 
<address@hidden>

signature.asc
Description: Digital signature