00:54 <icecream95> "(offset + sz) >= pool->base.slab_size". This is why most of the buffer overflows I found were never hit in normal usage... in effect the allocator always allocates an extra array element

00:54 <icecream95> (Should be ">" for slightly better memory usage at the expense of off-by-one errors doing Bad Things)

01:20 <icecream95> uhh... panfrost_set_stream_output_targets seems completely broken to me, why do both loops start at zero?

01:22 <icecream95> That will only work if every second invocatio of the function is with num_targets=0

03:07 <icecream95> The function is broken anyway; it doesn't handle NULL targets properly

10:23 <icecream95> The FUSE emulator runs *really* bad under Xwayland, but with Xorg it's almost perfect

10:23 <icecream95> I guess I should try to debug this to see what the problem is

10:25 <icecream95> Hmm... Xorg is using distro Mesa, Xwayland is using my fork

10:27 <icecream95> But Xwayland with distro Mesa still performs really badly