Re: svn commit: r956921 - /subversion/trunk/subversion/libsvn_fs_fs/structure

C. Michael Pilato wrote:
> Ooh... ouch. Yes, this is a bug. The lock setting code goes about things
> one way, and the lock deleting code expects another. I think it turns out
> okay because the lock system gracefully ignores missing lock files, but I
> further predict that it means that the serialized hash file for, say, the
> root of the repository, will always contain pointers to serialized hash
> files (which may or may not actually exist) for each and every file ever
> locked inside that repository. For a repository of a million files, that
> could eventually mean a million records -- perhaps all of which are dead --
> in that file. Forever.
>
> I'll dig around some to try to confirm this, and get an issue filed.

Tracking this in issue #3660 now:
http://subversion.tigris.org/issues/show_bug.cgi?id=3660

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand