Re: set_error_handler() and fatal errors?

> We'll have to address this in the documentation. Fatal errors cannot be > trapped in userland because the engine may not be in a stable state - and > we should bail out as quickly and as cleanly as possible. True, I was tracing through the call to undefined function case and didn't see any real reason not to be able to trap that. But other types of errors may be corruptively fatal. > We might want to create a new class of errors, errors which are fatal > unless they're trapped. I think that a call to an undefined function can > belong to such a class. I think that would be useful. There are probably a number of errors that are currently fatal that don't leave the engine in an unrecoverable state. -Rasmus