On 01.03.2015 15:23, Graham Leggett wrote: > On 27 Feb 2015, at 2:14 AM, Graham Leggett <minf...@sharp.fm> wrote: > >> By way of background, I added the regex named grouping support to httpd that >> allowed you to do stuff like <LocationMatch ^/svn/(?<CUSTOMERNAME>[^/]+)/>, >> and I’ve been adding expression support to a growing list of httpd >> directives, starting with all the require directives, and then following up >> with Alias and Redirect. >> >> What I’m trying to achieve is to lower the cost of httpd directives, both in >> terms of the amount of config required to achieve a certain aim, as well as >> making it possible to administer an httpd server (and svn repository) using >> changes to an LDAP directory only. >> >> I chose the two argument layout for SVNPath and SVNParentPath as it was both >> backwards compatible with existing configs and had the side effect that the >> expression could be rooted below a specific root directory for safety, given >> that the directory pointed at by SVNPath isn’t protected (to my knowledge) >> by <Directory> blocks. It also had the side effect that SVNParentPath could >> give us a list of repositories rather nicely. > Here is an updated patch below that fixes the return issue (apr_pstrcat does > nothing) and some formatting. > > I have found a new issue - mod_dav_svn (and mod_dav itself) don’t work > properly inside a <LocationMatch>, This is caused as the raw regex is passed > as the path, which then triggers an assertion failure inside libsvn. This > needs an httpd fix, will look at that in the mean time.
Hi Graham, You may have missed my mail from a couple days ago: > I think this would be a nice feature to have in mod_dav_svn. It is, IMO, > a better approach than the one currently implemented on the > SVNParentPathTemplate branch, even if it is limited to httpd-2.4+. > > Graham, I've created a branch for this: > > ^/subversion/branches/mod-dav-svn-expressions > > you can commit your patch there, if you will; you have the necessary > karma since Subversion-the-project uses the universal-commit bit. > > Once that's done, we'll want to tweak our test suite so that we can test > the new feature, and probably update docs and so on. Please go ahead and commit the patch (and any further changes) to the branch. -- Brane