Re: [Monotone-devel] segmentation fault in changesetify

[Nathaniel Smith]

On Tue, Apr 05, 2005 at 09:18:52AM +0200, Martin Kihlgren wrote:
> 
> Great that it explained the problem :)
> 
> Well, if you are willing to solve my problem, feel free to do it any
> way you like - all code in my database is currently quite stable, and
> I dont foresee any need to find any certain older revision other than
> for historic reasons...

Ah, well, then I'll go with the easiest solution.  Using your 0.17
monotone, on the post-migration database, try running:
  $ monotone --db=foo.db db execute "delete from manifest_certs where id = 
'5954cc14d6096749b9112cf59cce0b6ef304efe8' and name = 'ancestor'"
  $ monotone --db=foo.db db execute "delete from manifest_certs where id = 
'b7ba8f46ddcbd46b71127564469bdbb10bf8323f' and name = 'ancestor'"

This _should_ allow your changesetify to run, and leave you with a
working history.

It will break both loops, and leave you with graphs looking like

  A
  |
  B  C
  | /
  D
  |
  E

which may be a little mysterious; I'd suggest after you finish the
migration you use monotone-viz or similar to find the new revision ids
corresponding to the "C" node above (there will be two of them), and
using 'monotone comment' to attach some explanation of what happened,
so as not to confuse yourself when browsing history at some point in
the future...

> I would think the most elegant solution, though perhaps not that easy
> to fix, would be to patch the changesetify so that it solves this
> problem in one of the ways you mention below - but since this is
> perhaps a quite unusual problem and that solution does sound slightly
> complicated, I dont actually expect that to happen :)

This would indeed be elegant, but very difficult -- impossible in the
general case -- and generally not worth it for a command that you may
be the last person to ever run :-).  (Or if there are still other
pre-0.14 users holding out out there, at least there are only a small,
finite number of times this command remains to be run...)

Hope that helps,
-- Nathaniel

-- 
"But in Middle-earth, the distinct accusative case disappeared from
the speech of the Noldor (such things happen when you are busy
fighting Orcs, Balrogs, and Dragons)."