lvqcl wrote:
> So I wonder - does it makes sense to change this type too?
> The patch is attached.
I updated the tests to test unsigned as well as signed and then
applied your patch.
Cheers,
Erik
--
--
Erik de Castro Lopo
The patch
http://git.xiph.org/?p=flac.git;a=commitdiff;h=0bea5fb96436588b4e78c5be79bf174b9be7a202
changes the type of t value from FLAC__int32 to uint32_t,
but only for 24-bit signed input samples. 24-bit *un*signed
input still uses FLAC__int32 t.
So I wonder - does it makes sense to change
lvqcl wrote:
> The patch
> http://git.xiph.org/?p=flac.git;a=commitdiff;h=0bea5fb96436588b4e78c5be79bf174b9be7a202
> changes the type of t value from FLAC__int32 to uint32_t,
> but only for 24-bit signed input samples. 24-bit *un*signed
> input still uses FLAC__int32 t.
These UB fixes were
Erik de Castro Lopo wrote:
> These UB fixes were driven purely by UBSan warnings generated while running
> the test suite. The fact that no warning was gneerated by 24 bit unsigned
> input suggests that the test suite doesn't cover 24 bit unsigned input.
Yes, I can see only --sign=signed option