On Wed, Aug 6, 2014 at 7:21 AM, Branko Čibej <br...@wandisco.com> wrote:
> On 06.08.2014 07:19, Branko Čibej wrote: > > On 03.08.2014 17:41, stef...@apache.org wrote: > > Author: stefan2 > Date: Sun Aug 3 15:41:02 2014 > New Revision: 1615412 > > URL: http://svn.apache.org/r1615412 > Log: > * subversion/libsvn_subr/config.c > (svn_config_create2, > svn_config_get_config, > svn_config__get_default_config, > svn_config_addsection, > svn_config_copy_config): Use our custom hash function for faster > insertion and lookup. > > > Stefan, there's no reason to do this on the authzperf branch; it has very > little to do with authz performance, and it just clutters up the diff wrt. > trunk for no really good reason. > > If you really think this change is worthwhile, then it's OK to do it on > trunk –- *after* we branch 1.9. > > I made this change in conjunction of authz tuning. So, yes it is a general improvement and could have been made on /trunk directly if we weren't about to branch off 1.9, but it is not entirely unrelated to the purpose of the authzperf branch. > > Same for r1615419, IMO. > This can be merged to trunk independently from other changes. But concerning the initial commit going to this branch, I think this is a borderline case. It doubled authz parsing throughput. With your new data model, the relative gain of this change will eventually be even greater. -- Stefan^2.
