Re: assert infrastructure for runtime

[Mark Mitchell <mark at codesourcery dot com>]

Nathan Sidwell wrote:
> Mark Mitchell wrote:
>
> > Yes.  But, when adjusting the runtime, be careful to distinguish 
> > aborts that are really intended to be aborts from assertions!
>
>
> I'm confused. How is this different to adjusting the compiler?  Each
> abort is either 'we can't get here' (in which case it should be
> gcc_unreachable), or 'the condition protecting this is never true' (in
> which case it should be gcc_assert), or 'we've got scrogged input'
> (in which case some kind of error would be appropriate -- do we want a
> gcc_fatal for those?).

There might be situations where we can't print anything, but still need 
to abort.  Like std::terminate, for example, or perhaps things in the 
exception unwinder.  (You can't safely call "printf" during 
exception-unwinding because the exception might in a signal handler, and 
you can't safely call "printf" from a signal handler.)

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304