Bug 102924 - mesa (git version) images too dark
Summary: mesa (git version) images too dark
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: git
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Jason Ekstrand
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-09-21 10:11 UTC by Jos van Wolput
Modified: 2017-10-17 10:37 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Patch to try (3.68 KB, patch)
2017-09-28 04:21 UTC, Jason Ekstrand
Details | Splinter Review

Description Jos van Wolput 2017-09-21 10:11:21 UTC
Using the most recent git version of Mesa 17.3.0-devel (git-df8767a14e) all images are too dark, even with maximum brightness set.
Reverting to mesa-17.2.1 solves this issue.
My laptop is a HP Elitebook core i5 pro with Ivybridge processor.
Comment 1 Mark Janes 2017-09-21 17:38:23 UTC
Are you able to bisect this issue?  It would help us fix it.
Comment 2 Kenneth Graunke 2017-09-22 00:07:32 UTC
My assumption is that this is the sRGB bug caused by commit d80cbbeaff9329fdc78ae3d97097c1e65dfcdd61.  Jason I believe plans to work on the regression once everyone's back home from XDC and other conferences.

https://cgit.freedesktop.org/~kwg/mesa/commit/?h=gnome-srgb-woes has a potential workaround, but Jason decided that was not quite right.
Comment 3 Jos van Wolput 2017-09-22 09:29:38 UTC
(In reply to Mark Janes from comment #1)

Reverting src/mesa/drivers/dri/i965/intel_tex_image.c to the version of 
https://cgit.freedesktop.org/~kwg/mesa/commit/?h=gnome-srgb-woes
as mentioned by  Kenneth Graunke in Comment 2 fixes this issue.
Comment 4 Jason Ekstrand 2017-09-24 19:13:33 UTC
I know what needs to be done here but can't really get to it into Tuesday or Wednesday.
Comment 5 Jason Ekstrand 2017-09-28 04:21:45 UTC
Created attachment 134522 [details] [review]
Patch to try

Mind giving this version a try?  It should be the more "correct" fix as it doesn't whack the miptree format unnecessarily.
Comment 6 Kenneth Graunke 2017-09-28 04:54:30 UTC
Nope, that's even worse...it makes GNOME Shell crash as soon as I open a terminal.

(gdb) f 4
#4  0x00007fc7ef036da6 in intel_miptree_copy_slice (brw=0x5617e3f9e720, src_mt=
    0x5617e5be3b30, src_level=0, src_layer=0, dst_mt=0x5617e4f0aa00, 
    dst_level=0, dst_layer=0) at intel_mipmap_tree.c:1540
1540       assert(src_mt->format == dst_mt->format);
(gdb) bt
#0  0x00007fc8096d38a0 in raise () at /usr/lib/libc.so.6
#1  0x00007fc8096d4f09 in abort () at /usr/lib/libc.so.6
#2  0x00007fc8096cc0dc in __assert_fail_base () at /usr/lib/libc.so.6
#3  0x00007fc8096cc153 in  () at /usr/lib/libc.so.6
#4  0x00007fc7ef036da6 in intel_miptree_copy_slice (brw=0x5617e3f9e720, src_mt=
    0x5617e5be3b30, src_level=0, src_layer=0, dst_mt=
    0x5617e4f0aa00, dst_level=0, dst_layer=0) at intel_mipmap_tree.c:1540
#5  0x00007fc7ef0371ed in intel_miptree_copy_teximage (brw=0x5617e3f9e720, intelImage=0x5617e6db1040, dst_mt=0x5617e4f0aa00, invalidate=false)
    at intel_mipmap_tree.c:1625
#6  0x00007fc7ef048a85 in intel_finalize_mipmap_tree (brw=0x5617e3f9e720, unit=0) at intel_tex_validate.c:166
#7  0x00007fc7ef048bc1 in brw_validate_textures (brw=0x5617e3f9e720)
    at intel_tex_validate.c:198
#8  0x00007fc7eeffb148 in brw_try_draw_prims (ctx=0x5617e3f9e720, arrays=0x5617e3f86178, prims=0x7ffd90ad3c20, nr_prims=1, ib=0x0, index_bounds_valid=true, min_index=60, max_index=63, xfb_obj=0x0, stream=0, indirect=0x0) at brw_draw.c:627
#9  0x00007fc7eeffbc9c in brw_draw_prims (ctx=0x5617e3f9e720, prims=0x7ffd90ad3c20, nr_prims=1, ib=0x0, index_bounds_valid=1 '\001', min_index=60, max_index=63, gl_xfb_obj=0x0, stream=0, indirect=0x0) at brw_draw.c:856
#10 0x00007fc7eed00ff4 in vbo_draw_arrays (ctx=0x5617e3f9e720, mode=6, start=60, count=4, numInstances=1, baseInstance=0, drawID=0) at vbo/vbo_exec_array.c:486
#11 0x00007fc7eed017f8 in vbo_exec_DrawArrays (mode=6, start=60, count=4)
    at vbo/vbo_exec_array.c:641
#12 0x00007fc806e528a0 in  () at /usr/lib/mutter/libmutter-cogl-0.so
#13 0x00007fc806e96294 in  () at /usr/lib/mutter/libmutter-cogl-0.so
#14 0x00007fc806e96e1c in  () at /usr/lib/mutter/libmutter-cogl-0.so
#15 0x00007fc806e96596 in  () at /usr/lib/mutter/libmutter-cogl-0.so
#16 0x00007fc806e96843 in  () at /usr/lib/mutter/libmutter-cogl-0.so
#17 0x00007fc806e96596 in  () at /usr/lib/mutter/libmutter-cogl-0.so
#18 0x00007fc806e97baa in  () at /usr/lib/mutter/libmutter-cogl-0.so
#19 0x00007fc806e67349 in cogl_flush () at /usr/lib/mutter/libmutter-cogl-0.so
#20 0x00007fc806e9c7b5 in cogl_onscreen_swap_buffers_with_damage ()
    at /usr/lib/mutter/libmutter-cogl-0.so
#21 0x00007fc80a7c1204 in  () at /usr/lib/mutter/libmutter-clutter-0.so
#22 0x00007fc80a7c19ab in  () at /usr/lib/mutter/libmutter-clutter-0.so
#23 0x00007fc80a82ce47 in  () at /usr/lib/mutter/libmutter-clutter-0.so
#24 0x00007fc80a818002 in  () at /usr/lib/mutter/libmutter-clutter-0.so
#25 0x00007fc809cbea57 in g_main_context_dispatch ()
    at /usr/lib/libglib-2.0.so.0
#26 0x00007fc809cbec88 in  () at /usr/lib/libglib-2.0.so.0
#27 0x00007fc809cbefa2 in g_main_loop_run () at /usr/lib/libglib-2.0.so.0
#28 0x00007fc80b4840f0 in meta_run () at /usr/lib/libmutter-0.so.0
#29 0x00005617e3918f3c in main ()
Comment 7 Jos van Wolput 2017-10-10 11:44:40 UTC
(In reply to Jason Ekstrand from comment #5)
> Created attachment 134522 [details] [review] [review]
> Patch to try
> 
> Mind giving this version a try?  It should be the more "correct" fix as it
> doesn't whack the miptree format unnecessarily.
I applied your patch on mesa version 17.3.0-devel (git-ca949e00d8).
It works for me, the color is OK!
As I am not using GNOME Shell but XFCE4 I don't get a crash when I open a terminal.
Comment 8 Jos van Wolput 2017-10-17 10:37:06 UTC
Fixed by Kenneth Graunke (commit 9e779e59).
Thanks!


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.