During servicing a request, we are not careful enough to catch all
possible errors. In fact, a spurious, uncaught exception can bring
the whole server down.
1) We should investigate the reasons that a thrown error in a
server worker thread kills the app, not just the thread
2) The worker threads should be MUCH more robust, catching
anything and everything and logging if it ever gets that far
3) mod_render should catch anything and everything and produce
reasonable error pages, which can be very detailed if we are in
some debugging mode and less detailed if we are not. (Unclear
how such a mode is set...)
4) object.render() (a primitive) calls the tiny template expand()
routine. This routine can throw strings. object.render() should
catch them and give better error results
Logged In: YES
user_id=219202
1) the worker threads didn't catch - and if you throw out of a thread
stack - it kills the whole app
2) worker threads are now much more robust - they catch and syslog
anything that gets thrown as far as the thread's main loop
similar fixes were applied to the Checkpoint thread and the
MainServerLoop thread
Logged In: YES
user_id=219202
3) mod_render's catching of errors from execution was somewhat
improved -- it could still do better in the formatting department (this is
the only task left on this card)
4) object.render() was changed to catch the thrown strings from tiny
template