On 08/11/2006 07:12 PM, Nick Kew wrote: > On Friday 11 August 2006 17:01, [EMAIL PROTECTED] wrote: > > >>http://issues.apache.org/bugzilla/show_bug.cgi?id=40234 >> >> Summary: failed request with APR_INCOMPLETE >> Product: Apache httpd-2 >> Version: 2.3-HEAD >> >>On server/request.c(ap_directory_walk), >>call apr_stat with APR_FINFO_NAME. >> >> >>> if (!(opts & OPT_SYM_LINKS)) { >>> rv = apr_stat(&thisinfo, r->filename, >>> APR_FINFO_MIN | APR_FINFO_LINK, >>> r->pool); >>> if (rv != APR_SUCCESS) { > > > This looks like a regression from the fixes to the symlinks+mod_dir problem: > http://svn.apache.org/viewvc/httpd/httpd/trunk/server/request.c?r1=420983&r2=425394
Hm, to be honest I grabbed these flags from other code further down below and in line 958 of request.c we also use it. I for myself had no trouble with APR_FINFO_NAME during my tests of the patch on Linux. > > The reporter says the APR_FINFO_NAME is useless. That looks right > to me (we're only checking if it's a symlink). His patch makes sense on *X. Afer studying the APR documentation I agree that APR_FINFO_NAME could be removed. But as I am Unix centric I guess comments from non Unix people would be good. > > What about other platforms? Well, AFAICT it'll work. But it's also > irrelevant and could perhaps be optimised out? It was only > added to fix a Unix bug! > > Is this worth pursuing, or should we just apply the patch as-is? > Provided that this does not break things on other platforms I would just apply the patch as-is. Regards Rüdiger
