2008/10/21 Warren Layton <[EMAIL PROTECTED]>: > I'm trying to run cgi-bin/config.cgi from my browser but I keep > getting "500 Internal Server Error". When this error occurs, the > following message is appended to my Apache 2 error log: > > [Tue Oct 21 21:45:31 2008] [error] [client 127.0.0.1] Premature > end of script headers: config.cgi > > This is likely my fault and I'm assuming that I have misconfigured my > <Directory> section in /etc/apache2/sites-available/eg.conf. For > testing, I tried to "Allow from all": > > Alias /cgi-bin/ "/openils/var/cgi-bin/" > <Directory "/openils/var/cgi-bin"> > AddHandler cgi-script .cgi .pl > AllowOverride None > Options FollowSymLinks ExecCGI Indexes > Order allow,deny > Allow from all > </Directory> > > > ...but I still get the same error. > > Likely a silly misconfiguration of my own making, but just in case...
No, not at all. It turns out that this is a regression introduced in OpenSRF; if either the osrf_json_gateway or osrf_http_translator Apache modules are enabled, then CGI support in Apache is broken. Of course, both modules are required for Evergreen 1.4 to function. So - anyone want to start bisecting the changes in OpenSRF, testing just osrf_json_gateway and a minimalist CGI configuration to determine what commit introduced the regression? I can vouch that this was working as of May 2008, as I have a working system that dates back to that time. And if there are any testing fans out there with some time on their hands to contribute to the project, this would be a great opportunity to help set up a framework to help catch these problems early in the development cycle... -- Dan Scott Laurentian University
