httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@covalent.net>
Subject Re: cvs commit: httpd-2.0/server request.c
Date Tue, 16 Oct 2001 21:28:06 GMT
From: "Greg Ames" <gregames@remulak.net>
Sent: Tuesday, October 16, 2001 3:23 PM

> r->path_info definately has something to do with it.  
> 
> I just stepped thru handle_map_file & read_types_multi with yesterday
> morning's cvs, using our default /manual/ URI.  It finds the three
> language variants for index.html in that dir, and in my case picks
> index.html.en as best.  So far, so good.
> 
> At the very end of handle_map_file, it grabs r->path_info which is an
> empty string, massages it some yielding "/", concatenates the result
> with the "best" file name from the /manual/ directory (index.html.en)
> plus r->path_info once more (???) and does an internal redirect to
> /index.html.en .  bzzzzzzt - we just lost the /manual/ piece of the URI.

Ick... I'll look at this case.

> I'll cvs up to pick up this change, put my patch back on, & see what
> happens.

I'm absolutely -1 on 'escaping' from the <Location > restrictions when
we don't have to.  Relative paths to the current directory are one such
case.  This is new behavior in 2.0 (didn't test location_walk ever on file
subrequests in 1.3) but I see no reason why we don't do this properly.

Bill


Mime
View raw message