On Wed, Nov 25, 2009 at 07:28:22PM +0000, Philip Martin wrote: > "Jon Foster" <[email protected]> writes: > > > A patch is attached. The patch is against 1.6.6. I've tried to > > keep the changes minimal, because I think this might be a candidate > > for a Subversion 1.6.7 patch release. > > A minimal patch is something like this: > > Index: subversion/svnsync/main.c > =================================================================== > --- subversion/svnsync/main.c (revision 884213) > +++ subversion/svnsync/main.c (working copy) > @@ -299,7 +299,7 @@ > apr_sleep(apr_time_from_sec(1)); > } > } > - else > + else if (i != 9) > { > SVN_ERR(svn_ra_change_rev_prop(session, 0, SVNSYNC_PROP_LOCK, > mylocktoken, subpool)); > > > A #define to allow XXX-1 rather than 9 might be better.
What about the patch below? [[[ * subversion/svnsync/main.c (get_lock): The lock attempt is done in iteration N, and success is checked in iteration N+1. Prevent the case where svnsync attempts to take the lock in the very last iteration, doesn't check for success, and ends up erroring out, leaving behind a stale lock. Found by: Jon Foster <[email protected]> Based on patch by: philip ]]] Index: subversion/svnsync/main.c =================================================================== --- subversion/svnsync/main.c (revision 884790) +++ subversion/svnsync/main.c (working copy) @@ -276,7 +276,8 @@ get_lock(svn_ra_session_t *session, apr_pool_t *po subpool = svn_pool_create(pool); - for (i = 0; i < 10; ++i) +#define SVNSYNC_LOCK_RETRIES 10 + for (i = 0; i < SVNSYNC_LOCK_RETRIES; ++i) { svn_pool_clear(subpool); SVN_ERR(check_cancel(NULL)); @@ -299,8 +300,9 @@ get_lock(svn_ra_session_t *session, apr_pool_t *po apr_sleep(apr_time_from_sec(1)); } } - else + else if (i < SVNSYNC_LOCK_RETRIES - 1) { + /* Except in the very last iteration, try to set the lock. */ SVN_ERR(svn_ra_change_rev_prop(session, 0, SVNSYNC_PROP_LOCK, mylocktoken, subpool)); }
