CppCon 2017: John Regehr “Undefined Behavior in 2017 (part 1 of 2)”

[@annazolkieve9235]

Excellent informative talk describing details of the UB! Very useful! Many thanks!

[@gubgubbubbub]

Excellent speaker and content!

[@rafalmichalski4893]

Is it possible to have UB during compile time (assuming no compile errors) ?

[@hl2mukkel]

Fantastic talk, highly interesting!

[@Kobold666]

sizeof(int) is equal to sizeof(long) no matter if it's a 32bit or 64bit architecture. They changed int from 16bit to 32bit, but never to 64bit.

[@cocomonkilla]

Awesome lecture

[@thought2007]

9:12 - Why can't the Standard just say "this is implementation-defined"? The example explains UB vs. defined, but doesn't explain the advantage of undefined vs. implementation-defined.

[@Zeturic]

Because UB isn't just at the language level. An architecture could exist where shifting too far is undefined, even at the hardware level. By listing it as implementation-defined rather than undefined behavior, C/C++ implementations on that architecture would be forced to add to gracefully handle that case in software every time there's a shift.

[@BillyONeal]

The memcpy example is the standard being stupid. We should fix that.

[@BillyONeal]

_moonPtr 1. I mean it should be fixed in C. 2. Allowing more things doesn't break backcompat anyway.

[@omfgbunder2008]

Shouldn't 32bit

[@n3whous3]

Wouldn't be better to write the subtitles with somebody who is in the business? "Valgrint", j.f.c.