Re: input validation in C++ internal functions

[John W. Eaton]

I agree that Octave should never crash, even for internal functions 
called incorrectly.  But for internal functions, the quality of the 
error messages is not as critical.

For internal functions that are written in .m files, less error checking 
is needed because even for invalid scripting code, Octave shouldn't be 
crashing, though the error messages might be obscure.  Also, error 
checking in a small internal .m file function could significantly 
degrade performance.  That's not as big a problem in C++.

Ideally, all warnings and errors would have IDs associated with them so 
that you can control them.

If we want a comprehensive list of all error and warning IDs, is it 
possible to automate that process?  If not, then I think it would be a 
big maintenance problem.

Octave error and warning IDs should follow the format  Octave:foo-bar.

Should we automatically convert between known Matlab IDs that are 
frequently used and the corresponding Octave IDs when they exist?

jwe