Raymond Toy wrote:
> ...
>
> Perhaps we can make these signal errors that are handled to return
> NoObject things, but users/developers can make it so that the error is
> actually caught and throws the user to the debugger?
>
Many of these issues have been thrashed out in the context of the
handling of exceptions in IEEE floating-point.
Specifically when /how to control if/ the program should continue
running with NaNs or throw an exception.
Not everything that is done with NaNs is entirely relevant in the
symbolic context, but if you keep the differences
in mind, it is helpful to read what others have concluded. (A major
difference is that in the IEEE context and a
rigid memory limit (no growth) you have a strictly
limited number of bits (in double, 51) to record "what happened". Of
course in a system with memory allocation
a 51-bit pointer could be very useful, and that could be used in Lisp,
C, etc.)
> I've found it quite useful to have debugmode(true) when trying to figure
> out where my maxima routine was dividing by zero. It's really annoying
> to have some long running routine stop with a divide by zero error and
> no way of figuring out where that happened.
>
Does debugmode(true) really give you the info you need? Maybe there
should be a tutorial on this topic.
> Ray
>
>