On Tuesday, July 23, 2013 12:29:19 PM Ricardo Signes wrote: > We should apply a patch that does nothing but make internal errors become > objects with overloaded stringification. Then we can see how much of the > core tests, and of the CPAN, breaks. We may be surprised by the results. I worked on this for several days at the end of last year. My approach was simple: make Perl_croak_sv() turn exception strings into full-blown exception objects with the string as the payload. After some work faking up a new package ('CORE::Exception'), I flailed around with overload magic for a while, and never quite managed to get everything all working together to the point where enough of the test suite passed successfully that I could diagnose what went wrong. In retrospect, the worst part was probably trying to manage a new Perl class (well, namespace and methods and overloading) from XS. Loading it from a .pm file wasn't any better, as you can imagine that introduced introducing bootstrapping issues. I can't find the relevant branch at the moment, but I'm not sure there is much left salvaging at this point. -- cThread Previous | Thread Next