On Thu, Jan 16, 2014 at 02:33:23PM +0100, Olivier Berger wrote: > > AFAIU, the file may be touched before the first commit is made, so that > sqlite will not try to create it with wrong mask, and we are safe. This is > the workaround I'll suggest for fusionforge (see #735440). > > However, if repositories have received commits whereas the bug was already > present, then I fear some inconsistency if the perms are restored as writable > for committers, and some later commits are the only ones in the cache. > > I haven't found a mention of a tool that'd rebuild the cache, and the > removing the file just postpones the problem (it will be recreated, with > wrong mask, and will only contain the later commits). Note that the upstream > docs seem to be wrong in this respect > (http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_fs_fs/structure) > mentioning that "This file is only consulted during writes and never > during reads. Consequently, it is not required, and may be removed at an > abritrary time, with the subsequent loss of rep-sharing capabilities for > revisions written thereafter." > > Or maybe if the rep-cache.db isn't complete doesn't do harm ? > > I'll try and ask upstream for what to do for existing repositories that > exhibited the problem. >
Apparently, from reading http://mail-archives.apache.org/mod_mbox/subversion-users/201401.mbox/%3C52D83701.5030201%40reser.org%3E it seems that no real harm can happen if the file is deleted/recreated at some point. So we have a safe workaround in changing permissions at a later time, it seems. Oh, and btw, the bug in sqlite is at #608604. Hope this helps. Best regards, -- Olivier BERGER http://www-public.telecom-sudparis.eu/~berger_o/ - OpenPGP-Id: 2048R/5819D7E8 Ingenieur Recherche - Dept INF Institut Mines-Telecom, Telecom SudParis, Evry (France) -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
