[FFmpeg-devel] [PATCH] strict-aliasing-safe aes.c

[Måns Rullgård]

Michael Niedermayer <michaelni at gmx.at> writes:

> What i was saying is that code like *(uint32_t*)p shouldnt be a aliasing
> violation to begin with. The ones that are aliasing violations for a
> compiler should of course lead to erroring out.
> Warning about them and continuing with generating broken code is of course
> stupid

The spec is quite clear about which conditions shall be errors, and
aliasing violations are not among them.

-- 
M?ns Rullg?rd
mans at mansr.com