code browser should use oops system

Given that loggerhead is built on top of a completely different web framework, this is non-trivial.

Secondly, the way we trap timeouts in Launchpad (in the database adapter code) is not really appropriate for Loggerhead, since it doesn't issue database queries.

Leaving aside timeouts, is it really hard to put a traceback into the oops system? I understood that the recent cleanups by ddaa now mean it's separated from zope and can be just called with a traceback from any script..

(I guess you just said non-trivial, not "hard".)

The oops system is still dependent on the canonical.config system. Otherwise, it is not _really_ dependent on zope, for example it is used by the branch scanner which is a non-zope script.

Probably, the biggest problem here is to find a place in loggerhead to hook into to record exceptions. Details on how to report oopses from non-zope systems can be found on https://launchpad.canonical.com/ErrorReportManagementForScripts.

Now that loggerhead is wsgi based, I'm sure we can write some OOPS-reporting middleware.

I have a patch that adds a basic OOPS WSGI middleware to codebrowse. (I did most of it at the Epic, but didn't get around to making it functional until now.) Will put it in code review shortly.

The patch has landed, but still lacks the production config values. I've just mailed losas@ asking for what they should be.

Fixed in stable r11288 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/11288>.