On Wed, Jul 13, 2011 at 21:10, Daniel Shahaf <d...@daniel.shahaf.name> wrote: > Greg Stein wrote on Wed, Jul 13, 2011 at 20:54:56 -0400: >> On Wed, Jul 13, 2011 at 20:32, <danie...@apache.org> wrote: >> > Author: danielsh >> > Date: Thu Jul 14 00:32:05 2011 >> > New Revision: 1146528 >> > >> > URL: http://svn.apache.org/viewvc?rev=1146528&view=rev >> > Log: >> > * subversion/libsvn_fs_fs/fs_fs.c >> > (write_config): Document that 'svnadmin verify' will access (and thus >> > create) >> > rep-cache.db regardless of whether rep-sharing (of new reps) is enabled >> > in fsfs.conf. >> >> If rep-sharing is turned off, then it seems wrong to spontaneously >> create a .db that isn't begin used. > > It's harmless (it isn't being used). > > We could make the code pass readwrite (rather than rwcreate) for that > one caller, and then mask the error when the DB doesn't exist. (We do > want to verify the DB if it's exists but fsfs.conf says "Don't use it", > since fsfs.conf may get changed.) I'm not sure that non-creating an > unused db justifies this additional code.
I'm thinking of the sysadmin running into it, and thinking "wha...?! I turned that off. <goes to check>". Could be confusing. Cheers, -g
