As per irc conversation with lifeless, nzjrs and jelmer.
11:13 < lifeless> data looks completely fine
11:13 < lifeless> whats the branch you want to merge this with?
11:13 <@Jc2k> jstowers/conduit/devel on playground
11:13 <@Jc2k> which, aiui, is really just trunk right now
11:13 <@Jc2k> (which doesnt work either)
11:13 < nzjrs> I want to merge it with trunk. It hasnt been mreged in a week or so. The reason I started all this pain is because I want to push it too
gnome bzr now, for people to browse
11:14 < nzjrs> (what Jc2k said)
11:14 < lifeless> ok, lets see what the difference is; bzr missing http://bzr-playground.gnome.org/~jstowers/conduit/devel ...
11:15 < lifeless> I'm just doing a check first
11:16 < lifeless> hmm, inconsistent parents; I wonder if bzr-svn needs a tweak
11:16 < lifeless> and four ghosts
11:16 < lifeless> jelmer: ^
11:17 < lifeless> ok, different error
11:18 < lifeless> running bzr check in /bzr/jstoers/conduit/devel
11:19 < lifeless> I have to go. Jc2k - can you turn this into a bug (there is a root cause here somewhere; might be bzr-svn, might be bzr), mark it critical
(user behaviour should never be able to cause this sort of confusion)
11:20 < lifeless> 6 ghosts and 10 inconsistent parents
11:20 <@Jc2k> sure thing
11:20 < lifeless> hopefully jelmer or odd_bloke or someone in a eu or possibly us tz can look at this
11:21 < nzjrs> lifeless, can I continue working on these branches, just starting with fresh branches of each?
11:21 < lifeless> I'll be away for ~0 hours minimum
11:21 < jelmer> lifeless: bzr-svn doesn't push right hand side parent information for individual files
11:21 < lifeless> jelmer: that explains the ghosts perhaps; but not the inconsistent parents
11:21 < lifeless> jelmer: more concerning is the missing text errors
11:22 < nzjrs> lifeless, it might be a coincidence, but the last successful push of that branch to svn resulted in every file in svn getting touched
11:22 < nzjrs> which I thought was odd
11:22 <@Jc2k> yes, i've noticed that
11:22 <@Jc2k> its like it copied trunk on top of itself
11:22 <@Jc2k> (which confuses the bejesus out of the stock launchpad svn import iirc)
11:23 < lifeless> what is really strange i
11:23 < lifeless> *is*
11:23 < lifeless> that both branches appears to have the damaged revisions
11:23 < lifeless> but to have different bits of them
I strongly suspect bzr-svn at this point. comparing the actual inventory for one of the problem revisions between both trees is likely to show differences (which is a violation of the identity-per-key rule in bzr).