Boost C++ Libraries: Ticket #9236: Catch programmer errors with asserts, not exceptions

John Maddock — Fri, 11 Oct 2013 17:33:50 GMT

I believe this to be a true runtime error which should be checked and reported via an exception. Note that if you pass an unsigned integer as the exponent then the check is omitted, and in any case the cost of the check likely to be insignificant compared to the rest of the code.

There are however some inconsistencies in the way these checks are applied, which I'll fix shortly.

John Maddock — Sat, 12 Oct 2013 08:19:31 GMT

(In [86258]) Optimize copying of allocator-free cpp_int's via memcpy. Fix consistency of checks for exponents < 0 in powm. Fixes #9231. Refs #9236.

status changed; resolution set

John Maddock — Sat, 12 Oct 2013 08:20:10 GMT

status new → closed
resolution → wontfix

Domagoj Šarić — Sat, 12 Oct 2013 11:26:48 GMT

I wrote all those tickets in a bit of a hurry so I didn't check whether that one source code line I pasted is actually the one I ran into when stepping through your code
regardless, aren't exceptions for exceptional cases outside of a programmer's direct influence (network, filesystem, memory allocation,...)? IOW isn't an invalid input to a function a programmer error?
apply other comments on EH from #9234...

John Maddock — Sat, 12 Oct 2013 12:11:47 GMT

It's not a programmer error if the arguments come from user input (for example).