On 03/04/2016 03:51 AM, Mats Peterson wrote:
On 03/04/2016 03:49 AM, Michael Niedermayer wrote:
On Fri, Mar 04, 2016 at 03:41:23AM +0100, Mats Peterson wrote:
On 03/04/2016 03:30 AM, Michael Niedermayer wrote:
- if (!*palette && ret == CONTAINS_PAL)
- *palette = pkt->data + pkt->size - AVPALETTE_SIZE;
+ if (!*palette && ret == CONTAINS_PAL) {
+ uint8_t *pkt_pal = pkt->data + pkt->size - AVPALETTE_SIZE;
+ int i;
+ for (i = 0; i < AVPALETTE_COUNT; i++) {
+ uint8_t *p8 = pkt_pal + 4*i;
+ uint32_t *p32 = (uint32_t *)p8;
undefined behavior, violating alignment requirements, and possibly
aliassing violation
Really? It works just fine, so please elaborate.
you cant cast arbitrary pointers to uint32_t
uint32_t can have platform specific alignment requirements that are
stricter than what a random 4 uint8_t have.
pkt->size does not need to be a multiple of 4 so even if pkt->data
would be aligned enough p8 doesnt have to be
Whatever. Forget this version 6 of the patch set and use version 5, with
version 2 of patch 4/4 in it (the one I posted in PM as well). And I
would be grateful if you could apply it soon. It's extensively tested here.
Mats
I get what you mean with the alignment issues, though.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
