Working copy locks and client errors.

Hello

Earlier Greg Hudson asked why lock.c:pool_cleanup doesn't remove
physical locks from the working copy. It's because we may get there
when an error occurs running a log file, and removing a lock when a
log file is present is wrong.

Now attempting to run the log file is not sensible either (an error
may have occurred while running the log file...) but I suppose
pool_cleanup could check for the presence of a log file and remove the
lock if there is no log file. As far as I know that's the only thing
the lock protects at present.

Does this sound sensible? Is entry caching going to affect this
decision? What about the proposed solutions to issue 806? As far as
I can tell, so long as we maintain consistency between the tests used
by lock.c:pool_cleanup and svn_wc_cleanup (using a common function
say) this should be fine.

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org