RE: svnsync crash on empty repo
From: Bert Huijben <bert_at_qqmail.nl>
Date: Sat, 5 Sep 2015 18:45:16 +0200
> -----Original Message-----
An empty range isn't bad, but your patch appears to also explicitly allow undefined ranges, if I read it correctly.
I would call A range of r1 upto r0 defined (empty), but I would call a range or r4 upto r3 with current HEAD r3 undefined.
We use r1 and r0 as special in quite a few places where we really just want to point at the initial revision, I see no problem with that but I don't want to introduce a definition of 'HEAD+1' as there is no we can declare stable behavior on that... It may even change partway through an operation, the moment that revision is committed.
Svnrdump had a bug where it actually passed HEAD+1 as base revision of a commit in a specific case... It is better to catch these errors on the client than to rely on not having concurrent commits to define behavior of our tools.
This is an archived mail posted to the Subversion Dev mailing list.