14:00 <digetx> cyndis: what's the version of video player

14:01 <cyndis> 'mpv 0.33.1-dirty, from ALARM repository

14:01 <cyndis> +'

14:11 <digetx> 0.33.1 should work

14:11 <digetx> could you please apply https://dpaste.com/HWJQZQ2UH.txt

14:12 <digetx> maybe that's a pixman version difference

14:15 <digetx> the player log looks okay

14:17 <cyndis> surface_bitmap.c:191/vdp_bitmap_surface_put_bits_native(): pixman_blt failed x0=0 y0=0 width=979 height=112 source_pitches[0]=4096 pix_stride=4096 pfmt=537036936

14:32 <digetx> the parameters look good

14:44 <digetx> cyndis: do you see anything displayed at all? a black box?

14:45 <cyndis> I think there was a white line..I'll check again in a moment

14:59 <cyndis> the whole picture gets a tiny bit darker and there's a wide and vertically narrow box of garbage where the osd should be

14:59 <cyndis> the width looks about reasonable but it's only 10 pixels or so tall

15:00 <cyndis> the video itself does continue to play and be visible (but as said, a bit darker)

15:07 <digetx> that is very strange, it should be the only place where pixman_blt may fail https://cgit.freedesktop.org/pixman/tree/pixman/pixman-arm-neon.c#n231

15:07 <digetx> but the BPP should be okay

15:10 <cyndis> I can try to dig into it at some point

15:10 <digetx> thank you, will be great because I can't reproduce it

15:52 <cyndis> interesting. the PKGBUILD for pixman seems to disable all of the arm implementations

15:53 <cyndis> it looks like at some point in the past it only did that on ARMv6 and older (which seems appropriate since it looks like they all require some type of SIMD etc.) but now it is also for ARMv7

15:53 <cyndis> i wonder how it's possible no one else has run into this

15:57 <cyndis> actually, I think at least one backend should support ARMv6 as well *shrug*

16:06 <cyndis> digetx: yep, works now

16:07 <cyndis> there's still a slight hue change whenever the OSD is enabled, which is not there on my desktop

16:08 <cyndis> i wonder if there's some unaccounted for sRGB thing while OSD is being composited

16:14 <digetx> cyndis: do you know what is the reason of disabling arm implementations? shouldn't pixman fall back to a generic implementation in that case?

16:17 <cyndis> it doesn't seem to have any generic implementation

16:17 <cyndis> don't know why they would be disabled

16:18 <digetx> pixman-fast-path should be generic

16:19 <digetx> the hue change could be due to a missing CSC matrix support in kernel driver, the matrix used by gr2d should match to display or result may look differently when 2d blit is needed; the vdpau driver should fall back to 2d blits of each frame if default CSC is changed in player and kernel driver doesn't support CSC

16:21 <cyndis> hmm, yeah, don't know why fast-path is not being used

16:22 <cyndis> oh, I think fast-path doesn't set 'imp->blt'

16:23 <cyndis> which is what's used for pixman_blt

16:23 <digetx> yes

16:33 <digetx> but it works on T20 which doesn't have neon

16:35 <digetx> ah, there are neon and simd implementation

16:35 <digetx> could be worthwhile try to report it to arch maintainer, that is not good

16:36 <cyndis> yep, I'll look into it

16:47 <digetx> cool, thank you

19:39 <hexdump0815> thanks for setting up the logger