Trent Nelson wrote on Mon, Jul 09, 2012 at 23:17:28 -0700:
> On 7/8/12 11:48 AM, "Daniel Shahaf" <d.s_at_daniel.shahaf.name> wrote:
> >Daniel Shahaf wrote on Sat, Jul 07, 2012 at 15:09:38 +0100:
> >> How can we have a 'modify-file' within an added-without-copyfrom tree?
> That's a pretty impressive invariant violation. Enversion would have
> caught it, but only because we assert that a modify's previous path must
> match its current path -- not because we specifically look for this
> I've also never come across this in the wild. We should definitely ping
> `rmuir` to find out how he committed that revision; platform, version, etc.
> >>Isn't svnsync correct to complain in this case?
> >I'd like to resolve the situation on the live svn.a.o repository. I see
> >a couple of ways to do so:
> >- Hand-edit the revision file, changing "modify-file" to "add-file "
> >- Eliminate r1356317 from history: create an svnsync copy of /repos/asf
> > using an authz file that excludes
> > (or maybe just
> > and have the Lucene PMC recreate that tag later
> I'd lean towards a scripted version of option 1. You could then
> disseminate the script to svnsync mirror admins (like myself), perhaps via
> an ASF/infrastructure blog post?
There won't be a need for that: recent version of svnsync refuse to sync
r1356317 because of the invariant violation (and I imagine anyone who
mirrors the ASF repos uses a recent svnsync), so only svn-master.a.o has
the corruption. (Even the EU mirror doesn't have it.)
Pending rmuir@'s feedback, then, I'll go ahead and edit the revision
file in-place. No change should be needed to any of the svn mirrors.
 Note that no svn*.apache.org DNS name resolves to the EU mirror
right now --- not even svn.eu.a.o.
Received on 2012-07-10 10:47:31 CEST