> > On 06/21/2017 10:41 AM, Christophe Fergeau wrote: > > On Tue, Jun 20, 2017 at 08:39:44PM +0300, Uri Lublin wrote: > >> On 06/20/2017 04:52 PM, Frediano Ziglio wrote: > >>> memmove already deal with any alignment so there's no > >>> reason to have row byte pointer cast to uint32_t. > >> > >> Hi Frediano, > >> > >> I think the code has no problem with memmove itself, but > >> with the pointer used later being not-aligned (but only for > >> non-x86 arch ?). > > > > "later" ? The only uses of dest_aligned/dest_misaligned in this function > > are as args to memmove, for which alignment does not matter. > > With "later" I meant the callers of canvas_fix_alignment, and > likely "later" pixman. > The image is going to get accessed by pixman that probably > expects rows to be aligned, which is why the bits are memmove'd. > > SPICE_ALIGNED_CAST can verify (if configured for it) that the > pointer is indeed aligned. > > Uri. > The image is allocated by pixman itself in the way pixman want it. The code is there to fix alignment from network which has no padding for rows. Frediano _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel