www-apache-bugdb mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Link-M Webmaster <webmas...@link-m.de>
Subject Re: apache-api/6885: PATH_TRANSLATED generation produces garbage in error_log
Date Sat, 02 Dec 2000 16:20:00 GMT
The following reply was made to PR apache-api/6885; it has been noted by GNATS.

From: Link-M Webmaster <webmaster@link-m.de>
To: Tony Finch <dot@dotat.at>
Cc: apbugs@apache.org, Michael Buschbeck <m.buschbeck@buero.link-m.de>
Subject: Re: apache-api/6885: PATH_TRANSLATED generation produces garbage in error_log
Date: Sat, 2 Dec 2000 17:19:38 +0100

      Hi there!
 On 01.12.00 (00:48), Tony Finch wrote:
 > >Linux 2.2.17 #1 Mit Nov 15 13:38:42 CET 2000 i586 unknown w/
 > >apache 1.3.9 compiled by gcc; in newer apache releases'
 > >CHANGES files I couldn't find anything related to the bug as
 > >described below, so I haven't upgraded first.
 > Did you try with older versions to verify that this behaviour
 > was introduced in 1.3.9?
 No. In fact, I'd expected to see the problem around for some
 > >When apache encounters an URI such as
 > >/cgi-bin/myscript.cgi/path/to/template.html it issues a
 > >subrequest to itself in order to get PATH_TRANSLATED. In
 > >combination with access restrictions on the file
 > >"/path/to/template.html" (e.g. a FilesMatch block anywhere in
 > >the server configuration), this behaviour leads to lots of
 > >unwanted error_log messages, as every access to the script is
 > >logged as a failed access to the raw template file.
 > I think this is a misfeature rather than a bug,
 > since it would be hard to fix -- there isn't a single location
 > in the code where Apache could decide not to log access
 > control errors because this is a subrequest or whatever; to
 > the contrary, the logging is scattered all over the place.
 On the other hand: Why is a 'real' subrequest necessary in order
 to get PATH_TRANSLATED?  Innocent proposal: Why not use some
 kind of 'request simulator' that might get used by other pieces
 of the code as well?  Would this have a chance to make those
 routines faster oder would it rather hurt apache's performance?
 > I'm also not convinced that this is a big problem in practice.
 Well, it clutters our error_log's with some hundreds of
 kilobytes of garbage per month, thus making it harder to find
 real error information when needed.
      Best regards,
      Philipp Rotmann
 Linksystem Muenchen GmbH                     webmaster@link-m.de
 Schloerstrasse 10                           http://www.link-m.de
 80634 Muenchen                              Tel. 089 / 890 518-0
 We make the Net work.                       Fax 089 / 890 518-77

View raw message