On 02.06.2015 18:30, Bert Huijben wrote: > Strange… this should never be necessarily when *removing* something > from a hash (value =NULL). > > I think you see some kind of other problem. > > Bert > > Sent from Surface > > *From:* Philip Martin <mailto:phi...@apache.org> > *Sent:* Tuesday, June 2, 2015 5:19 PM > *To:* comm...@subversion.apache.org <mailto:comm...@subversion.apache.org> > > Author: philip > Date: Tue Jun 2 15:19:26 2015 > New Revision: 1683126 > > URL: http://svn.apache.org/r1683126 > Log: > Avoid using pool memory after clear in FSX. > > * subversion/libsvn_fs_x/cached_data.c > (read_dir_entries): Copy hashed values into hash pool. > > Modified: > subversion/trunk/subversion/libsvn_fs_x/cached_data.c > > Modified: subversion/trunk/subversion/libsvn_fs_x/cached_data.c > URL: > http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs_x/cached_data.c?rev=1683126&r1=1683125&r2=1683126&view=diff > ============================================================================== > --- subversion/trunk/subversion/libsvn_fs_x/cached_data.c (original) > +++ subversion/trunk/subversion/libsvn_fs_x/cached_data.c Tue Jun 2 > 15:19:26 2015 > @@ -2494,7 +2494,9 @@ read_dir_entries(apr_array_header_t *ent > { > /* We must be in incremental mode */ > assert(hash); > - apr_hash_set(hash, entry.key, entry.keylen, NULL); > + apr_hash_set(hash, > + apr_pstrmemdup(scratch_pool, entry.key, > entry.keylen), > + entry.keylen, NULL); > continue; > }
I agree with Bert: this change is unnecessary. -- Brane
