[<<][packetforth][>>][..]
Wed Nov 29 10:29:35 EST 2006

error reporting and backtrace handling in task context

now, what should a backtrace be? probably a restartable
task. unfortunately, that's not always possible.. but, maybe it's a
good idea to try to make it always possible. vm->current needs to be
valid at all times, so restarting is possible. the primitive doesn't
really need to be saved. the xt is more important.

so what about making backtrace capture explicit? in other words: if
you want debugging, you need to start a word in a different task. that
would take all the backtrace stuff out of the core. 

the next problem is 'clear-error' and the way errors are registered.

ok.. currently i think it's sort of ok. the ad-hoc error message stuff
is moved to a separate word, and i have a restartable backtrace using
tasks, with all internal backtrace code removed (throw will just drop
stuff).



[Reply][About]
[<<][packetforth][>>][..]