On 11/05/15 13:57, Markus Armbruster wrote: > Laszlo Ersek <ler...@redhat.com> writes: > >> On 11/04/15 17:35, Gabriel L. Somlo wrote: > [...] >>> + assert(size > 0 && size <= sizeof(value)); >> >> It's a matter of taste, and I won't insist at all, just mention that I >> didn't write those two assert()s as separate statements :) >> >> Namely, with a conjunction (P1 && P2 && ... && Pn), you have the >> possibility to spell the assertion as: >> >> assert(P1); >> assert(P2); >> ... >> assert(Pn); >> >> And, if any one of those fails, you will know *which one*. Because the >> line number in the "assertion failed" message will tell you. > > Yes, matter of taste, but that can't stop me having opinions on matters > of taste ;) > > You pay for the more detailed assertion failure reporting with extra > source code clutter (as written, it's immediately obvious that it's a > bounds check, less so if split), and extra object code when NDEBUG is > off (which it should always be). > > Personally, I'm content to fish details out of a core dump. YMMV.
The 12 GB core dump that the Launchpad user, reporting the bug from the other side of the Earth, failed to save? :) But, I do concede your point. (Line numbers aren't a panacea either, in situations like the above.) Gabriel, please pick whichever format you like more. Thanks Laszlo
