Conversation

Notices

1. Embed this notice
   Andy Price (andyprice@mastodon.social)'s status on Saturday, 16-Nov-2024 07:02:34 JST Andy Price

   Undefined behaviour in a programming language spec is a good thing because it enables optimisations. Similarly, taking the windscreen wipers off your car is a good thing because it improves aerodynamics.

   • Embed this notice
     Andy Price (andyprice@mastodon.social)'s status on Saturday, 16-Nov-2024 07:02:32 JST Andy Price

     in reply to

     • The Penguin of Evil

     @etchedpixels Sequence points is an interesting case because the unsequenced operations are effectively defined by omission :) You still don't get compiler warnings about something like foo(&s) + bar(&s) though so I'm sure there are a million bugs out there due to it. Is that worth it, is the question.

   • Embed this notice
     pinskia (pinskia@hachyderm.io)'s status on Saturday, 16-Nov-2024 07:02:32 JST pinskia

     in reply to

     • The Penguin of Evil

     @andyprice @etchedpixels `foo(&s) + bar(&s)` is a little more obvious an issue rather than say `foo(bar1(), bar2())`. And yes that has shown up in real code (GCC in fact).
     See https://gcc.gnu.org/PR12411 and
     http://gcc.gnu.org/ml/gcc-patches/2003-09/msg01545.html

   • Embed this notice
     The Penguin of Evil (etchedpixels@mastodon.social)'s status on Saturday, 16-Nov-2024 07:02:33 JST The Penguin of Evil

     in reply to

     • Yukari Hafner :v_lesbian:

     @shinmera @andyprice Used sensibly a lot of them (like C sequence points) make an enormous performance difference to code.

   • Embed this notice
     Yukari Hafner :v_lesbian: (shinmera@mastodon.tymoon.eu)'s status on Saturday, 16-Nov-2024 07:02:34 JST Yukari Hafner :v_lesbian:

     in reply to

     @andyprice UB always struck me less as being for optimisations, but rather for getting whoever was funding the standard to sign off on it because they had an existing product they couldn't be bothered to make comply