Feargal Reilly wrote:
That's debatable. While it does indicate a programming error,
that shouldn't take down the server. It should make sure the
database is returned to a clean state and that the current action
fails with a significant error. Given that it is a programming
error, realistically it will only be triggered by some bizarre
bug, caused by something like being handed a message which ends
with\r\r\n\rr\/\n'\r\n\r, or the digits of the remote user's IP
address adding up to 42.
It should certainly scream in all the relevant logs, but I don't
think taking the server down is appropiate.
Having said that, having a hook to force a failure would be
useful in testing and some non-critical deployments.
Makes sense.
Perhaps surround it by #ifdef DEBUG, and only let if fail if we're in
the DEBUG section.
OTOH, when it really fails (takes down the server) we will probably
notice any problems a lot earlier.
Ilja
