Thread began on users@. Implentation thoughts in the email, so copying to
dev@.
Any people replying: Please restrict the recipients according to the content
of your message.
kfogel@collab.net wrote:
> Roland Schwingel <Roland.Schwingel@onevision.de> writes:
>> Is there any news about this bug in the past 3 weeks? It appears to be my
>> sole killer to migrate from CVS to SVN.
>>
>> Thanks again for your help,
>
> Max Bowsher has been working on cvs2svn bugs as fast as he can (which
> turns out to be pretty fast!). I don't know whether he plans to work
> on this one -- or if so, when -- but you could try debugging it
> yourself or asking Max if there's anything you can do to help him,
> such as come up with a smaller reproduction case, write a regression
> test, etc, etc.
I need to polish off the rather large number of almost-complete issues I've
accumulated, but after that, I will have a look at 1440. BTW, I think 1427
is at least partially related, and fixing 1440 may be closely tied up with
fixing 1471.
Current situation:
Branches out of dead revisions are handled such that it works most of the
time, but can both crash, or produce incorrect results when exposed to
certain unusual RCS files.
I can see an easy fix which would avoid the above problems:
Treat "resurrection across branches"[*1] as a simple add (no history). This
probably isn't a such bad idea, as we do not attempt any special handling
for "resurrection within a branch"[*2].
*1 = Delete, make branch, re-add on the new branch
*2 = Delete, later re-add.
The downside, is, of course, that with this change we output less copyfrom
history than we currently do. I think that's better than crashing and
possible erroneous behaviour though.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Sep 23 00:17:36 2003