[Bug 795391] New: vaapi: problems when playing with glimagesink with egl

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Bug ID: 795391
Summary: vaapi: problems when playing with glimagesink with egl
Classification: Platform
Product: GStreamer
Version: git master
OS: All
Status: NEW
Severity: normal
Priority: Normal
Component: gstreamer-vaapi
Assignee: gstreamer-***@lists.freedesktop.org
Reporter: ***@igalia.com
QA Contact: gstreamer-***@lists.freedesktop.org
CC: ***@gmail.com, ***@igalia.com
GNOME version: ---

Here are problems I found.

$ gst-play-1.0 a.mp4 b.mp4 c.mp4 --videosink=glimagesink

1. Crash on Wayland.
Not 100% but a crash is very likely to happen when jumping to next media.

2. Doesn't display anything from the 2nd media.
You can see this issue on X11/EGL and Wayland(if the 1st one is fixed).

If you can't reproduce these issues, please let me know because it means that
my environment is wrong.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #1 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Confirmed.

There's a problem when dmabuf is negotiated. Perhaps we're not resetting
something properly.

The crash I got in wayland (weston) is:

#0 0x00007ffff649de7b in __GI_raise (sig=***@entry=0x6) at
../sysdeps/unix/sysv/linux/raise.c:51
#1 0x00007ffff649f231 in __GI_abort () at abort.c:79
#2 0x00007ffff64969da in __assert_fail_base (fmt=0x7ffff65e9d48 "%s%s%s:%u:
%s%sAssertion `%s' failed.\n
%n", assertion=***@entry=0x7fffcf8d5831 "batch->buffer->virtual",
file=***@entry=0x7fffcf8d5630 ".
./subprojects/intel-vaapi-driver/src/intel_batchbuffer.c",
line=***@entry=0x3f, function=***@entry=
0x7fffcf8d5a60 <__PRETTY_FUNCTION__.10416> "intel_batchbuffer_reset") at
assert.c:92
#3 0x00007ffff6496a52 in __GI___assert_fail
(assertion=***@entry=0x7fffcf8d5831 "batch->buffer->vi
rtual", file=***@entry=0x7fffcf8d5630
"../subprojects/intel-vaapi-driver/src/intel_batchbuffer.c", line=
***@entry=0x3f, function=***@entry=0x7fffcf8d5a60
<__PRETTY_FUNCTION__.10416> "intel_batchbuffer_re
set") at assert.c:101
#4 0x00007fffcf6d903a in intel_batchbuffer_reset
(batch=***@entry=0x7fff8005cea0, buffer_size=0x80000)
at ../subprojects/intel-vaapi-driver/src/intel_batchbuffer.c:63
#5 0x00007fffcf6d91f5 in intel_batchbuffer_flush
(batch=***@entry=0x7fff8005cea0)
at ../subprojects/intel-vaapi-driver/src/intel_batchbuffer.c:148
#6 0x00007fffcf6bff4e in gen6_pp_object_walker (pp_context=0x7fff8007af40,
ctx=<optimized out>)
at ../subprojects/intel-vaapi-driver/src/i965_post_processing.c:4674
#7 0x00007fffcf6bff4e in gen6_pp_pipeline_setup (pp_context=0x7fff8007af40,
ctx=<optimized out>)
at ../subprojects/intel-vaapi-driver/src/i965_post_processing.c:4691
#8 0x00007fffcf6bff4e in gen6_post_processing (ctx=<optimized out>,
pp_context=0x7fff8007af40, src_surfa
ce=<optimized out>, src_rect=<optimized out>, dst_surface=<optimized out>,
dst_rect=<optimized out>, pp_i
ndex=<optimized out>, filter_param=<optimized out>)
at ../subprojects/intel-vaapi-driver/src/i965_post_processing.c:4719
#9 0x00007fffcf6c4293 in i965_post_processing_internal (filter_param=0x0,
pp_index=0xe, dst_rect=0x7fff9
4f61e68, dst_surface=0x7fff94f61e80, src_rect=0x7fff94f61e60,
src_surface=0x7fff94f61e70, pp_context=<opt
imized out>, ctx=0x7fff80059990) at
../subprojects/intel-vaapi-driver/src/i965_post_processing.c:4743
#10 0x00007fffcf6c4293 in i965_image_pl1_rgbx_processing
(ctx=***@entry=0x7fff80059990, src_surface=src_s
***@entry=0x7fff94f61e70, src_rect=0x7fff94f61e60,
dst_surface=***@entry=0x7fff94f61e80, dst_r
ect=***@entry=0x7fff94f61e68) at
../subprojects/intel-vaapi-driver/src/i965_post_processing.c:5073
#11 0x00007fffcf6c66dc in i965_image_processing (ctx=***@entry=0x7fff80059990,
src_surface=***@en
try=0x7fff94f61e70, src_rect=***@entry=0x7fff94f61e60,
dst_surface=***@entry=0x7fff94f61e80,
dst_rect=***@entry=0x7fff94f61e68)
at ../subprojects/intel-vaapi-driver/src/i965_post_processing.c:5555
#12 0x00007fffcf6a83b0 in i965_hw_putimage (dst_rect=0x7fff94f61e68,
src_rect=0x7fff94f61e60, obj_image=0
x7fff8007a0a0, obj_surface=0x7fff80078f80, ctx=0x7fff80059990)
at ../subprojects/intel-vaapi-driver/src/i965_drv_video.c:5569
#13 0x00007fffcf6a83b0 in i965_PutImage (ctx=0x7fff80059990, surface=0x4000000,
image=0xa000000, src_x=0x
0, src_y=<optimized out>, src_width=0x40, src_height=0x40, dest_x=0x0,
dest_y=0x0, dest_width=0x40, dest_
height=0x40) at ../subprojects/intel-vaapi-driver/src/i965_drv_video.c:5630
#14 0x00007fffdd838f7c in gst_vaapi_surface_put_image
(surface=***@entry=0x555555927450, image=image@
entry=0x7fff80017170) at
../subprojects/gstreamer-vaapi/gst-libs/gst/vaapi/gstvaapisurface.c:741
#15 0x00007fffdd809dea in ensure_allowed_raw_caps (plugin=0x7fff8001a0f0,
***@entry=0x7fffd0060b60)
at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapipluginbase.c:1315
#16 0x00007fffdd809dea in gst_vaapi_plugin_base_get_allowed_raw_caps
(plugin=***@entry=0x7fff8001a0f0)
at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapipluginbase.c:1351
#17 0x00007fffdd80ecca in ensure_allowed_sinkpad_caps (postproc=0x7fff8001a0f0)
at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapipostproc.c:1095
#18 0x00007fffdd80ecca in gst_vaapipostproc_transform_caps_impl
(direction=<optimized out>, trans=0x7fff8
001a0f0) at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapipostproc.c:1189
#19 0x00007fffdd80ecca in gst_vaapipostproc_transform_caps
(trans=0x7fff8001a0f0, direction=<optimized ou
t>, caps=0x5555559274a0, filter=0x7fffd00302d0)
at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapipostproc.c:1212
#20 0x00007ffff62353de in gst_base_transform_transform_caps
(trans=***@entry=0x7fff8001a0f0, direction$
GST_PAD_SRC, caps=***@entry=0x5555559274a0,
filter=***@entry=0x7fffd00302d0)
---Type <return> to continue, or q <return> to quit---
at ../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:465
#21 0x00007ffff623a358 in gst_base_transform_query_caps (filter=0x7fffd00302d0,
pad=0x7fffd00cae70, tran$
=0x7fff8001a0f0) at
../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:686
#22 0x00007ffff623a358 in gst_base_transform_default_query
(trans=0x7fff8001a0f0, direction=<optimized o$
t>, query=0x7fffc8002320) at
../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:1541
#23 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x7fffd00cae70,
query=***@entry=0x7fffc8002320)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#24 0x00007ffff7653c82 in gst_pad_peer_query (pad=***@entry=0x7fffd00ca780,
query=***@entry=0x7fffc800$
320) at ../subprojects/gstreamer/gst/gstpad.c:4170
#25 0x00007ffff768e544 in gst_pad_peer_query_caps
(pad=***@entry=0x7fffd00ca780, filter=***@entry=0x7$
ffd00302d0) at ../subprojects/gstreamer/gst/gstutils.c:3089
#26 0x00007ffff623a25d in gst_base_transform_query_caps (filter=0x7fffc8002190,
pad=0x7fffd00ca9d0, tran$
=0x7fffd002f180) at
../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:669
#27 0x00007ffff623a25d in gst_base_transform_default_query
(trans=0x7fffd002f180, direction=<optimized o$
t>, query=0x7fffc8002280) at
../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:1541
#28 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x7fffd00ca9d0,
query=***@entry=0x7fffc8002280)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#29 0x00007ffff7653c82 in gst_pad_peer_query (pad=***@entry=0x7fffd0045a00,
query=0x7fffc8002280)
at ../subprojects/gstreamer/gst/gstpad.c:4170
#30 0x00007ffff7688d98 in query_caps_func (pad=***@entry=0x7fffd0045a00,
data=***@entry=0x7fff94f62540)
at ../subprojects/gstreamer/gst/gstutils.c:2759
#31 0x00007ffff765226e in gst_pad_forward (pad=***@entry=0x7fffd0045c50,
forward=***@entry=0x7ffff768
8d80 <query_caps_func>, user_data=***@entry=0x7fff94f62540)
at ../subprojects/gstreamer/gst/gstpad.c:3009
#32 0x00007ffff768b7ea in gst_pad_proxy_query_caps
(pad=***@entry=0x7fffd0045c50, query=***@entry=0x7ff
fc8002280) at ../subprojects/gstreamer/gst/gstutils.c:2809
#33 0x00007ffff7652570 in gst_pad_query_caps_default (query=0x7fffc8002280,
pad=0x7fffd0045c50)
at ../subprojects/gstreamer/gst/gstpad.c:3188
#34 0x00007ffff7652570 in gst_pad_query_default (pad=0x7fffd0045c50,
parent=<optimized out>, query=0x7fff
c8002280) at ../subprojects/gstreamer/gst/gstpad.c:3420
#35 0x00007fffdfdd8cbd in gst_queue_handle_sink_query (pad=<optimized out>,
parent=0x7fffc001b360, query=
0x7fffc8002280) at ../subprojects/gstreamer/plugins/elements/gstqueue.c:1079
#36 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x7fffd0045c50,
query=***@entry=0x7fffc8002280)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#37 0x00007ffff7653c82 in gst_pad_peer_query (pad=***@entry=0x7fffd0045310,
query=***@entry=0x7fffc8002
280) at ../subprojects/gstreamer/gst/gstpad.c:4170
#38 0x00007ffff768e544 in gst_pad_peer_query_caps
(pad=***@entry=0x7fffd0045310, filter=***@entry=0x7f
ffc8002190) at ../subprojects/gstreamer/gst/gstutils.c:3089
#39 0x00007ffff7945f49 in __gst_video_element_proxy_getcaps
(element=***@entry=0x7fff80018fb0, sinkpa
d=<optimized out>, srcpad=0x7fffd0045310, initial_caps=<optimized out>,
filter=***@entry=0x7fff80002ca
0) at
../subprojects/gst-plugins-base/gst-libs/gst/video/gstvideoutilsprivate.c:108
#40 0x00007ffff7935eb9 in gst_video_decoder_proxy_getcaps
(decoder=***@entry=0x7fff80018fb0, caps=<op
timized out>, filter=***@entry=0x7fff80002ca0)
at
../subprojects/gst-plugins-base/gst-libs/gst/video/gstvideodecoder.c:1750
#41 0x00007fffdd805572 in gst_vaapidecode_sink_getcaps (vdec=<optimized out>,
filter=<optimized out>)
at ../subprojects/gstreamer-vaapi/gst/vaapi/gstvaapidecode.c:1340
#42 0x00007ffff7935f7b in gst_video_decoder_sink_getcaps (filter=<optimized
out>, decoder=0x7fff80018fb0)
at ../subprojects/gst-plugins-base/gst-libs/gst/video/gstvideodecoder.c:1764
#43 0x00007ffff7935f7b in gst_video_decoder_sink_query_default
(decoder=0x7fff80018fb0, query=0x7fffc8002
2d0) at
../subprojects/gst-plugins-base/gst-libs/gst/video/gstvideodecoder.c:1813
#44 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x7fffd0044530,
query=***@entry=0x7fffc80022d0)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#45 0x00007ffff768bf94 in gst_pad_query_caps (pad=0x7fffd0044530,
filter=0x7fff80002ca0)
---Type <return> to continue, or q <return> to quit---
at ../subprojects/gstreamer/gst/gstutils.c:3043
#46 0x00007ffff793603b in gst_video_decoder_sink_query_default
(decoder=0x7fff80018fb0, query=0x7fffc8002
000) at
../subprojects/gst-plugins-base/gst-libs/gst/video/gstvideodecoder.c:1837
#47 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x7fffd0044530,
query=***@entry=0x7fffc8002000)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#48 0x00007ffff7653c82 in gst_pad_peer_query (pad=***@entry=0x7fffd0450cf0,
query=0x7fffc8002000)
at ../subprojects/gstreamer/gst/gstpad.c:4170
#49 0x00007ffff7688d60 in query_accept_caps_func (pad=***@entry=0x7fffd0450cf0,
data=***@entry=0x7fff94f
62c50) at ../subprojects/gstreamer/gst/gstutils.c:2699
#50 0x00007ffff765226e in gst_pad_forward (pad=***@entry=0x5555559d1a60,
forward=***@entry=0x7ffff768
8d50 <query_accept_caps_func>, user_data=***@entry=0x7fff94f62c50)
at ../subprojects/gstreamer/gst/gstpad.c:3009
#51 0x00007ffff768b65e in gst_pad_proxy_query_accept_caps
(pad=***@entry=0x5555559d1a60, query=***@entr
y=0x7fffc8002000) at ../subprojects/gstreamer/gst/gstutils.c:2739
#52 0x00007ffff7652945 in gst_pad_query_accept_caps_default
(query=0x7fffc8002000, pad=0x5555559d1a60)
at ../subprojects/gstreamer/gst/gstpad.c:3135
#53 0x00007ffff7652945 in gst_pad_query_default (pad=0x5555559d1a60,
parent=<optimized out>, query=0x7fff
c8002000) at ../subprojects/gstreamer/gst/gstpad.c:3416
#54 0x00007ffff7653688 in gst_pad_query (pad=***@entry=0x5555559d1a60,
query=***@entry=0x7fffc8002000)
at ../subprojects/gstreamer/gst/gstpad.c:4038
#55 0x00007ffff768e6fc in gst_pad_query_accept_caps
(pad=***@entry=0x5555559d1a60, caps=<optimized out>)
at ../subprojects/gstreamer/gst/gstutils.c:3126
#56 0x00007ffff764bb8a in pre_eventfunc_check (event=0x7fffd00a8670,
pad=0x5555559d1a60)
at ../subprojects/gstreamer/gst/gstpad.c:5572
#57 0x00007ffff764bb8a in gst_pad_send_event_unchecked
(pad=***@entry=0x5555559d1a60, event=***@entry=0
x7fffd00a8670, type=<optimized out>) at
../subprojects/gstreamer/gst/gstpad.c:5706
#58 0x00007ffff76569b1 in gst_pad_send_event (pad=0x5555559d1a60,
event=0x7fffd00a8670)
at ../subprojects/gstreamer/gst/gstpad.c:5885
#59 0x00007ffff4930128 in send_sticky_event (pad=<optimized out>,
event=<optimized out>, user_data=0x7fff
94f63030) at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1969
#60 0x00007ffff764a9b4 in foreach_dispatch_function
(pad=***@entry=0x7fffd819bc10, ev=***@entry=0x7fff94f6
2f30, user_data=***@entry=0x7fff94f62f80) at
../subprojects/gstreamer/gst/gstpad.c:5984
#61 0x00007ffff764a458 in events_foreach (pad=***@entry=0x7fffd819bc10,
func=***@entry=0x7ffff764a980 <f
oreach_dispatch_function>, user_data=***@entry=0x7fff94f62f80)
at ../subprojects/gstreamer/gst/gstpad.c:611
#62 0x00007ffff7657503 in gst_pad_sticky_events_foreach
(pad=***@entry=0x7fffd819bc10, foreach_func=forea
***@entry=0x7ffff4930110 <send_sticky_event>,
user_data=***@entry=0x7fff94f63030)
at ../subprojects/gstreamer/gst/gstpad.c:6015
#63 0x00007ffff492e0e9 in send_sticky_events (dbin=0x5555559e6060,
pad=0x7fffd819bc10)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1984
#64 0x00007ffff492e0e9 in connect_pad (dbin=***@entry=0x5555559e6060,
src=***@entry=0x7fffd002ee40, dpad
=***@entry=0x7fffd40a6ae0, pad=***@entry=0x7fffd819bc10,
caps=***@entry=0x7fff80002ca0, factories=0x555
55592a140, chain=0x7fffd05422f0, deadend_details=0x7fff94f630f8)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:2506
#65 0x00007ffff49390c4 in analyze_new_pad (dbin=***@entry=0x5555559e6060,
src=***@entry=0x7fffd002ee40,
pad=***@entry=0x7fffd819bc10, caps=***@entry=0x7fff80002ca0, chain=<optimized
out>,
***@entry=0x7fffd05422f0, new_chain=***@entry=0x7fff94f63148)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1798
#66 0x00007ffff4939a14 in pad_added_cb (element=***@entry=0x7fffd002ee40,
pad=***@entry=0x7fffd819bc1
0, chain=***@entry=0x7fffd05422f0)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:2939
#67 0x00007ffff4939c31 in caps_notify_cb (pad=0x7fffd819bc10, unused=<optimized
out>, chain=0x7fffd05422f
0) at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:3129
---Type <return> to continue, or q <return> to quit---
#68 0x00007ffff6bc5f6d in g_closure_invoke () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#69 0x00007ffff6bd8d3e in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#70 0x00007ffff6be13f5 in g_signal_emit_valist () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#71 0x00007ffff6be1e0f in g_signal_emit () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#72 0x00007ffff6bca424 in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#73 0x00007ffff76089d4 in gst_object_dispatch_properties_changed
(object=0x7fffd819bc10, n_pspecs=0x1, ps
pecs=0x7fff94f636e0) at ../subprojects/gstreamer/gst/gstobject.c:430
#74 0x00007ffff6bcc969 in g_object_notify_by_pspec () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#75 0x00007ffff764a8cc in store_sticky_event (pad=***@entry=0x7fffd819bc10,
event=0x7fffd00a8670)
at ../subprojects/gstreamer/gst/gstpad.c:5197
#76 0x00007ffff7656650 in gst_pad_push_event (pad=***@entry=0x7fffd819bc10,
event=<optimized out>)
at ../subprojects/gstreamer/gst/gstpad.c:5490
#77 0x00007ffff6239473 in gst_pad_set_caps (caps=0x7fff80002ca0,
pad=0x7fffd819bc10)
at ../subprojects/gstreamer/gst/gstcompat.h:59
#78 0x00007ffff6239473 in gst_base_transform_setcaps
(trans=***@entry=0x7fffd002ee40, pad=<optimized ou
t>, incaps=<optimized out>) at
../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:1328
#79 0x00007ffff623a98d in gst_base_transform_sink_eventfunc
(trans=***@entry=0x7fffd002ee40, event=0x7f
ffc4003a70) at ../subprojects/gstreamer/libs/gst/base/gstbasetransform.c:1889
#80 0x00007fffdfda96b1 in gst_capsfilter_sink_event (trans=0x7fffd002ee40,
event=0x7fffc4003a70)
at ../subprojects/gstreamer/plugins/elements/gstcapsfilter.c:521
#81 0x00007ffff764bd67 in gst_pad_send_event_unchecked
(pad=***@entry=0x7fffd0044090, event=***@entry=0
x7fffc4003a70, type=<optimized out>,
***@entry=GST_PAD_PROBE_TYPE_EVENT_DOWNSTREAM)
at ../subprojects/gstreamer/gst/gstpad.c:5715
#82 0x00007ffff764c1a7 in gst_pad_push_event_unchecked
(pad=***@entry=0x7fffd819b770, in_event=***@e
ntry=0x7fff94f63ab8, type=***@entry=GST_PAD_PROBE_TYPE_EVENT_DOWNSTREAM)
at ../subprojects/gstreamer/gst/gstpad.c:5368
#83 0x00007ffff764c618 in push_sticky (pad=***@entry=0x7fffd819b770,
ev=***@entry=0x7fff94f63b00, user_dat
a=***@entry=0x7fff94f63b60) at ../subprojects/gstreamer/gst/gstpad.c:3891
#84 0x00007ffff764a458 in events_foreach (pad=***@entry=0x7fffd819b770,
func=***@entry=0x7ffff764c560 <p
ush_sticky>, user_data=***@entry=0x7fff94f63b60) at
../subprojects/gstreamer/gst/gstpad.c:611
#85 0x00007ffff7656385 in check_sticky (event=<optimized out>,
pad=0x7fffd819b770)
at ../subprojects/gstreamer/gst/gstpad.c:3951
#86 0x00007ffff7656385 in gst_pad_push_event (pad=***@entry=0x7fffd819b770,
event=<optimized out>)
at ../subprojects/gstreamer/gst/gstpad.c:5502
#87 0x00007fffddcff3a2 in gst_pad_set_caps (caps=0x7fff80002ca0,
pad=0x7fffd819b770)
at ../subprojects/gstreamer/gst/gstcompat.h:59
#88 0x00007fffddcff3a2 in gst_h264_parse_update_src_caps
(h264parse=***@entry=0x7fff8001acf0, caps=
<optimized out>, ***@entry=0x7fff7c0040f0)
at ../subprojects/gst-plugins-bad/gst/videoparsers/gsth264parse.c:1991
#89 0x00007fffddd02097 in gst_h264_parse_set_caps (parse=<optimized out>,
caps=<optimized out>)
at ../subprojects/gst-plugins-bad/gst/videoparsers/gsth264parse.c:2726
#90 0x00007ffff621bb0c in gst_base_parse_sink_event_default
(parse=0x7fff8001acf0, event=0x7fffc40040e0)
at ../subprojects/gstreamer/libs/gst/base/gstbaseparse.c:1153
#91 0x00007fffddd017c3 in gst_h264_parse_event (parse=0x7fff8001acf0,
event=0x7fffc40040e0)
at ../subprojects/gst-plugins-bad/gst/videoparsers/gsth264parse.c:2905
#92 0x00007ffff764bd67 in gst_pad_send_event_unchecked
(pad=***@entry=0x7fffd819b9c0, event=***@entry=0
x7fffc40040e0, type=<optimized out>) at
../subprojects/gstreamer/gst/gstpad.c:5715
#93 0x00007ffff76569b1 in gst_pad_send_event (pad=0x7fffd819b9c0,
event=0x7fffc40040e0)
at ../subprojects/gstreamer/gst/gstpad.c:5885
#94 0x00007ffff4930128 in send_sticky_event (pad=<optimized out>,
event=<optimized out>, user_data=0x7fff
94f641d0) at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1969
#95 0x00007ffff764a9b4 in foreach_dispatch_function
(pad=***@entry=0x7fffd819b2d0, ev=***@entry=0x7fff94f6
40d0, user_data=***@entry=0x7fff94f64120) at
../subprojects/gstreamer/gst/gstpad.c:5984
---Type <return> to continue, or q <return> to quit---
#96 0x00007ffff764a458 in events_foreach (pad=***@entry=0x7fffd819b2d0,
func=***@entry=0x7ffff764a980 <f
oreach_dispatch_function>, user_data=***@entry=0x7fff94f64120)
at ../subprojects/gstreamer/gst/gstpad.c:611
#97 0x00007ffff7657503 in gst_pad_sticky_events_foreach
(pad=***@entry=0x7fffd819b2d0, foreach_func=forea
***@entry=0x7ffff4930110 <send_sticky_event>,
user_data=***@entry=0x7fff94f641d0)
at ../subprojects/gstreamer/gst/gstpad.c:6015
#98 0x00007ffff492e0e9 in send_sticky_events (dbin=0x5555559e6060,
pad=0x7fffd819b2d0)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1984
#99 0x00007ffff492e0e9 in connect_pad (dbin=***@entry=0x5555559e6060,
src=0x7fffd001a730,
***@entry=0x7fff7c00f6b0, dpad=***@entry=0x7fffd40a6ae0,
pad=0x7fffd819b2d0,
***@entry=0x7fffd819ae30, caps=***@entry=0x7fff7c0040f0,
factories=0x7fffc000b9a0, chain=0x7fffd0542
2f0, deadend_details=0x7fff94f64298)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:2506
#100 0x00007ffff49390c4 in analyze_new_pad (dbin=***@entry=0x5555559e6060,
src=***@entry=0x7fff7c00f6b0,
pad=***@entry=0x7fffd819ae30, caps=***@entry=0x7fff7c0040f0, chain=<optimized
out>,
***@entry=0x7fffd0215830, new_chain=***@entry=0x7fff94f642e8)
at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:1798
#101 0x00007ffff4939a14 in pad_added_cb (element=0x7fff7c00f6b0,
pad=0x7fffd819ae30, chain=0x7fffd0215830
) at ../subprojects/gst-plugins-base/gst/playback/gstdecodebin2.c:2939
#102 0x00007ffff52bdfce in ffi_call_unix64 () at
/usr/lib/x86_64-linux-gnu/libffi.so.6
#103 0x00007ffff52bd93f in ffi_call () at /usr/lib/x86_64-linux-gnu/libffi.so.6
#104 0x00007ffff6bc6761 in g_cclosure_marshal_generic ()
at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#105 0x00007ffff6bc5f6d in g_closure_invoke () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#106 0x00007ffff6bd8d3e in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#107 0x00007ffff6be13f5 in g_signal_emit_valist () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#108 0x00007ffff6be1e0f in g_signal_emit () at
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
#109 0x00007ffff7630568 in gst_element_add_pad
(element=***@entry=0x7fff7c00f6b0, pad=0x7fffd819ae30)
at ../subprojects/gstreamer/gst/gstelement.c:715
#110 0x00007fff8638137c in gst_qtdemux_add_stream (list=0x7fffd4002c00,
stream=0x7fff80013810, qtdemux=0x
7fff7c00f6b0) at ../subprojects/gst-plugins-good/gst/isomp4/qtdemux.c:8311
#111 0x00007fff8638137c in qtdemux_expose_streams
(qtdemux=***@entry=0x7fff7c00f6b0)
at ../subprojects/gst-plugins-good/gst/isomp4/qtdemux.c:12094
#112 0x00007fff8639ca0d in gst_qtdemux_loop_state_header
(qtdemux=***@entry=0x7fff7c00f6b0)
at ../subprojects/gst-plugins-good/gst/isomp4/qtdemux.c:4493
#113 0x00007fff863a0028 in gst_qtdemux_loop (pad=0x7fffd00442e0)
at ../subprojects/gst-plugins-good/gst/isomp4/qtdemux.c:6138
#114 0x00007ffff7680209 in gst_task_func (task=0x7fffd0220a70)
at ../subprojects/gstreamer/gst/gsttask.c:332
#115 0x00007ffff6e7e7d0 in () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#116 0x00007ffff6e7de05 in () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#117 0x00007ffff5fdd5aa in start_thread (arg=0x7fff94f65700) at
pthread_create.c:463
#118 0x00007ffff655fcbf in clone () at
../sysdeps/unix/sysv/linux/x86_64/clone.S:95

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #2 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #1)

Post by "GStreamer" (GNOME Bugzilla)
Confirmed.
There's a problem when dmabuf is negotiated. Perhaps we're not resetting
something properly.

Thanks for confirmation.
I have a patch for the crash, and now I'm thinking about the solution for the
2nd problem.

I'm going to propose some patches.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #3 from Hyunjun Ko <***@igalia.com> ---
BTW, I guess this is a cause of the issue
https://bugs.webkit.org/show_bug.cgi?id=184574

But I'm not sure yet.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #4 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #1)

Post by "GStreamer" (GNOME Bugzilla)
Confirmed.
There's a problem when dmabuf is negotiated. Perhaps we're not resetting
something properly.

Weird.
That's different from mine.
Mine is the following:

#0 0x00007ffff63d4fd2 in __memmove_avx_unaligned_erms () at
../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:407
#1 0x00007fffda193dd5 in memcpy (__len=<optimised out>, __src=<optimised out>,
__dest=0x0) at /usr/include/x86_64-linux-gnu/bits/string_fortified.h:34
#2 0x00007fffda193dd5 in gen9_render_init (ctx=<optimised out>) at
gen9_render.c:1832
#3 0x00007fffda164a9e in i965_Init (ctx=0x7fffa0023fa0) at
i965_drv_video.c:7513
#4 0x00007fffda164a9e in __vaDriverInit_1_1 (ctx=0x7fffa0023fa0) at
i965_drv_video.c:7678
#5 0x00007fffe2f03579 in va_openDriver (dpy=***@entry=0x7fffa0022be0,
driver_name=***@entry=0x7fffa017fd70 "i965") at va.c:447
#6 0x00007fffe2f046eb in vaInitialize (dpy=***@entry=0x7fffa0022be0,
major_version=***@entry=0x7fffc0c5b840,
minor_version=***@entry=0x7fffc0c5b844) at va.c:720
#7 0x00007ffff0634ee8 in vaapi_initialize (dpy=0x7fffa0022be0) at
gstvaapiutils.c:112
#8 0x00007ffff062a24e in gst_vaapi_display_create_unlocked (data=<optimised
out>, init_type=<optimised out>, display=0x5555559a0f40) at
gstvaapidisplay.c:896
#9 0x00007ffff062a24e in gst_vaapi_display_create (init_value=<optimised out>,
init_type=<optimised out>, display=0x5555559a0f40) at gstvaapidisplay.c:909
#10 0x00007ffff062a24e in gst_vaapi_display_new (display=0x5555559a0f40,
init_type=<optimised out>, init_value=<optimised out>) at
gstvaapidisplay.c:1155
#11 0x00007ffff05ffbd4 in gst_vaapi_create_display_from_handle
(handle=<optimised out>, display_type=<optimised out>) at
gstvaapipluginutil.c:126
#12 0x00007ffff05ffbd4 in gst_vaapi_create_display_from_gl_context
(gl_context_object=<optimised out>) at gstvaapipluginutil.c:200
#13 0x00007ffff05ffbd4 in gst_vaapi_ensure_display (element=0x7fffa017f090,
type=***@entry=GST_VAAPI_DISPLAY_TYPE_ANY) at gstvaapipluginutil.c:293
#14 0x00007ffff05fda1e in gst_vaapi_plugin_base_ensure_display
(plugin=***@entry=0x7fffa017f090) at gstvaapipluginbase.c:415
#15 0x00007ffff06018e1 in gst_vaapipostproc_ensure_display
(postproc=0x7fffa017f090) at gstvaapipostproc.c:219
#16 0x00007ffff06018e1 in gst_vaapipostproc_ensure_filter
(postproc=0x7fffa017f090) at gstvaapipostproc.c:228
#17 0x00007ffff0604136 in gst_vaapipostproc_start (trans=<optimised out>) at
gstvaapipostproc.c:310
#18 0x00007ffff6024e10 in gst_base_transform_activate
(trans=***@entry=0x7fffa017f090, active=***@entry=1) at
gstbasetransform.c:2390
#19 0x00007ffff60250a5 in gst_base_transform_sink_activate_mode (pad=<optimised
out>, parent=0x7fffa017f090, mode=<optimised out>, active=1) at
gstbasetransform.c:2456
#20 0x00007ffff718e16b in activate_mode_internal (pad=0x7fffe4110f00,
parent=0x7fffa017f090, mode=GST_PAD_MODE_PUSH, active=1) at gstpad.c:1224
#21 0x00007ffff718ebe6 in gst_pad_set_active (pad=***@entry=0x7fffe4110f00,
active=1) at gstpad.c:1107
#22 0x00007ffff716c66d in activate_pads (vpad=<optimised out>,
ret=0x7fffc0c5bb10, active=0x7fffc0c5bb6c) at gstelement.c:3040

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #5 from Hyunjun Ko <***@igalia.com> ---
Created attachment 371313
--> https://bugzilla.gnome.org/attachment.cgi?id=371313&action=edit
libs: display: avoid VA initialization twice

Since the commit dcf135e2 landed, it tries to initalize VA display
twice in case of creating GstVaapiDisplayEGL.

This is fine on X11 at least but leads to crash in the driver on Wayland.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #6 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
(In reply to Hyunjun Ko from comment #5)

Created attachment 371313 [details] [review]
libs: display: avoid VA initialization twice
Since the commit dcf135e2 landed, it tries to initalize VA display
twice in case of creating GstVaapiDisplayEGL.
This is fine on X11 at least but leads to crash in the driver on Wayland.

I would rather to resurrect the parenthood logic (and document it). It is
clearer than the proposed semantics in this patch.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #7 from Hyunjun Ko <***@igalia.com> ---
Created attachment 371315
--> https://bugzilla.gnome.org/attachment.cgi?id=371315&action=edit
RFC: display: egl: fix to create VaapiDisplayEGL with native EGL display and
call it when creation

gst_vaapi_display_egl_new_with_native_display has been broken since no
one uses it.

Currently it needs to call this api to create display with provided EGL
display so that it could avoid duplicated calls for the native display
(eg. eglTerminate).

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #8 from Hyunjun Ko <***@igalia.com> ---
The 2nd problems happens because gst-vaapi calls eglTerminate even though the
egl display is still used in gst-gl.

I'm requesting comments for the patch so that we could find better solution.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #9 from Hyunjun Ko <***@igalia.com> ---
Created attachment 371316
--> https://bugzilla.gnome.org/attachment.cgi?id=371316&action=edit
libs: utils: egl: mark is_wraaped TRUE if the context is wrapped

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371316|none |accepted-commit_now
status| |

--- Comment #10 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Review of attachment 371316:
--> (https://bugzilla.gnome.org/review?bug=795391&attachment=371316)

lgtm

we should also backport it to 1.14 and 1.12

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371315|none |reviewed
status| |

--- Comment #11 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Review of attachment 371315:
--> (https://bugzilla.gnome.org/review?bug=795391&attachment=371315)

::: gst-libs/gst/vaapi/gstvaapidisplay_egl.c
@@ +51,3 @@
guint display_type;
guint gles_version;
+ gboolean is_native;

why not just add gpointer egl_display instead of juggling with different
semantics in one single variable?

@@ +117,3 @@
#endif
#if USE_WAYLAND
+ if (!native_vaapi_display)

we could use params->display_type to choose what vaapi display instantiates,
besides the compilation dependencies

::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);

why not call here too gst_gl_display_get_handle() thus you could get a common
guintptr variable, which could be null by default.

@@ +259,3 @@
+
+ if (gl_display)
+ gst_object_unref (gl_display);

I don't see the reason for this unref since gl_display was unrefed already.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371313|none |rejected
status| |

--- Comment #12 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Review of attachment 371313:
--> (https://bugzilla.gnome.org/review?bug=795391&attachment=371313)

I'm going to mark this as rejected, because it's just a proof-of-concept.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Julien Isorce <***@gmail.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@gmail.com

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #13 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #6)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Hyunjun Ko from comment #5)

Created attachment 371313 [details] [review] [review]
libs: display: avoid VA initialization twice
Since the commit dcf135e2 landed, it tries to initalize VA display
twice in case of creating GstVaapiDisplayEGL.
This is fine on X11 at least but leads to crash in the driver on Wayland.

I would rather to resurrect the parenthood logic (and document it). It is
clearer than the proposed semantics in this patch.

I don't understand what you mean.

The parent thing is related to display-cache, which we removed. See commit
ec3e10f.

Which one do you want to resurrect?

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #14 from Hyunjun Ko <***@igalia.com> ---
Review of attachment 371315:
--> (https://bugzilla.gnome.org/review?bug=795391&attachment=371315)

::: gst-libs/gst/vaapi/gstvaapidisplay_egl.c
@@ +51,3 @@
guint display_type;
guint gles_version;
+ gboolean is_native;

Well, that's what I thought for a while.
Adding another gpointer is fine to me, too.

@@ +118,3 @@
#if USE_WAYLAND
+ if (!native_vaapi_display)
+ native_vaapi_display = gst_vaapi_display_wayland_new (NULL);

This is not about this patch, isn't it?
Anyway it still needs "#if USE_WAYLAND" since the header is included under the
macro.

::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);

Ah i didn't realize it. Thanks.

@@ +259,3 @@
+
+ if (gl_display)
+ gst_object_unref (gl_display);

That unref of gl_display is removed in this patch :)

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #15 from Hyunjun Ko <***@igalia.com> ---
(In reply to Hyunjun Ko from comment #14)

Post by "GStreamer" (GNOME Bugzilla)
::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);
Ah i didn't realize it. Thanks.

Hmm, Victor, as far as I understand, it has to get GLDisplayEGL so that it
retrieve EGLDisplay finally from it.

Could you explain what you mean in detail please?

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #16 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 371373
--> https://bugzilla.gnome.org/attachment.cgi?id=371373&action=edit
libs: egl: utils: fix usage of GstGL macros

Include gl.h for the required GstGL symbols.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371373|none |accepted-commit_now
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371373|accepted-commit_now |none
status| |
Attachment #371373|0 |1
is obsolete| |

--- Comment #17 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 371389
--> https://bugzilla.gnome.org/attachment.cgi?id=371389&action=edit
libs: egl: utils: fix usage of GstGL macros

Include gl.h for the required GstGL symbols.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #18 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 371390
--> https://bugzilla.gnome.org/attachment.cgi?id=371390&action=edit
libs: egl: utils: mark context as wrapped when it is

The returning egl context may be null, so we should check the
return value.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #19 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 371391
--> https://bugzilla.gnome.org/attachment.cgi?id=371391&action=edit
meson: fix USE_GLES_VERSION_MASK

1. The macro in the code is USE_GLES_VERSION_MASK
2. glesv3 is provided by glesv2 pkg-config, then it's required to
check headers

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371316|0 |1
is obsolete| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #20 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Attachment 371389 pushed as 9fde93f - libs: egl: utils: fix usage of GstGL
macros
Attachment 371390 pushed as 4af46f0 - libs: egl: utils: mark context as wrapped
when it is
Attachment 371391 pushed as 785efdb - meson: fix USE_GLES_VERSION_MASK

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371390|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371391|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371389|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #21 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
branch 1.14

* 56b63720 meson: fix USE_GLES_VERSION_MASK
* 5de03007 libs: egl: utils: mark context as wrapped when it is
* b50e4225 libs: egl: utils: fix usage of GstGL macros

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #22 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
branch 1.12

* cf67d6d5 meson: fix USE_GLES_VERSION_MASK
* 1660bd0d libs: egl: utils: mark context as wrapped when it is

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #23 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Review of attachment 371315:
--> (https://bugzilla.gnome.org/review?bug=795391&attachment=371315)

By the way, this patch will need to be rebased after bug 793643 is merged

::: gst-libs/gst/vaapi/gstvaapidisplay_egl.c
@@ +118,3 @@
#if USE_WAYLAND
+ if (!native_vaapi_display)
+ native_vaapi_display = gst_vaapi_display_wayland_new (NULL);

yes, we still need the macro. Just we could create efficiently the internal
display by identifying the environment instead of try-and-error

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #24 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #23)

Post by "GStreamer" (GNOME Bugzilla)
By the way, this patch will need to be rebased after bug 793643 is merged

Okay, I'll be working on it after those patches are merged

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #25 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 371465
--> https://bugzilla.gnome.org/attachment.cgi?id=371465&action=edit
plugins: handle EGL when creating VAAPI display from gl

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #26 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #25)

Created attachment 371465 [details] [review]
plugins: handle EGL when creating VAAPI display from gl

I cook this patch to create a EGL VAAPI display when the gl display is EGL. But
it brings problems when playing a second video, as those expressed here.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #27 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #23)

Post by "GStreamer" (GNOME Bugzilla)
By the way, this patch will need to be rebased after bug 793643 is merged
::: gst-libs/gst/vaapi/gstvaapidisplay_egl.c
@@ +118,3 @@
#if USE_WAYLAND
+ if (!native_vaapi_display)
+ native_vaapi_display = gst_vaapi_display_wayland_new (NULL);
yes, we still need the macro. Just we could create efficiently the internal
display by identifying the environment instead of try-and-error

I see, but in case of GST_VAAPI_DISPLAY_TYPE_EGL, we have to do try-and-error.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #28 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
(In reply to Hyunjun Ko from comment #27)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Víctor Manuel Jáquez Leal from comment #23)

Post by "GStreamer" (GNOME Bugzilla)
By the way, this patch will need to be rebased after bug 793643 is merged
::: gst-libs/gst/vaapi/gstvaapidisplay_egl.c
@@ +118,3 @@
#if USE_WAYLAND
+ if (!native_vaapi_display)
+ native_vaapi_display = gst_vaapi_display_wayland_new (NULL);
yes, we still need the macro. Just we could create efficiently the internal
display by identifying the environment instead of try-and-error

I see, but in case of GST_VAAPI_DISPLAY_TYPE_EGL, we have to do
try-and-error.

Agree. That's the case with attachment 371465

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #29 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #11)

Post by "GStreamer" (GNOME Bugzilla)
::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);
why not call here too gst_gl_display_get_handle() thus you could get a
common guintptr variable, which could be null by default.

There are different cases.
If there's a gldisplaywayland in glcontextegl, in this case, we should call
gst_gl_display_egl_from_gl_display to retrieve EGLDisplay.
(This is the case that's using gst-play or totem.)

But if gldisplayegl in glcontextegl, we can use just gst_gl_display_get_handle.
(This is the case that's in WebkitGTK.)

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #30 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
(In reply to Hyunjun Ko from comment #29)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Víctor Manuel Jáquez Leal from comment #11)

Post by "GStreamer" (GNOME Bugzilla)
::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);
why not call here too gst_gl_display_get_handle() thus you could get a
common guintptr variable, which could be null by default.

There are different cases.
If there's a gldisplaywayland in glcontextegl, in this case, we should call
gst_gl_display_egl_from_gl_display to retrieve EGLDisplay.
(This is the case that's using gst-play or totem.)
But if gldisplayegl in glcontextegl, we can use just
gst_gl_display_get_handle.
(This is the case that's in WebkitGTK.)

You can see my patch in attachment 371465 and use the variable display_type to
decide which API call use.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Michael Catanzaro <***@gnome.org> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@gnome.org
See Also| |https://bugs.webkit.org/sho
| |w_bug.cgi?id=184574

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #31 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #30)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Hyunjun Ko from comment #29)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Víctor Manuel Jáquez Leal from comment #11)

Post by "GStreamer" (GNOME Bugzilla)
::: gst/vaapi/gstvaapipluginutil.c
@@ +211,3 @@
+ GstGLDisplayEGL *egl_display;
+
+ egl_display = gst_gl_display_egl_from_gl_display (gl_display);
why not call here too gst_gl_display_get_handle() thus you could get a
common guintptr variable, which could be null by default.

There are different cases.
If there's a gldisplaywayland in glcontextegl, in this case, we should call
gst_gl_display_egl_from_gl_display to retrieve EGLDisplay.
(This is the case that's using gst-play or totem.)
But if gldisplayegl in glcontextegl, we can use just
gst_gl_display_get_handle.
(This is the case that's in WebkitGTK.)

You can see my patch in attachment 371465 [details] [review] and use the
variable display_type to decide which API call use.

hmm.. that wouldn't be simple imo.
(just using gst_gl_display_egl_from_gl_display VS using 2 apis by conditional
statement)

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #33 from Hyunjun Ko <***@igalia.com> ---
(In reply to Víctor Manuel Jáquez Leal from comment #32)

Comment on attachment 371315 [details] [review]
RFC: display: egl: fix to create VaapiDisplayEGL with native EGL display and
call it when creation
ping?
@hyunjun: what's the status of this patch?

I have no plan on this for now.
As we discussed before, it needs refactroing, doesn't it?
If so, you'd better do that than me.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #34 from Hyunjun Ko <***@igalia.com> ---
(In reply to Hyunjun Ko from comment #33)

Post by "GStreamer" (GNOME Bugzilla)
(In reply to Víctor Manuel Jáquez Leal from comment #32)

Comment on attachment 371315 [details] [review] [review]
RFC: display: egl: fix to create VaapiDisplayEGL with native EGL display and
call it when creation
ping?
@hyunjun: what's the status of this patch?

I have no plan on this for now.
As we discussed before, it needs refactroing, doesn't it?
If so, you'd better do that than me.

Ah you talked about the 2nd patch. Sorry about misunderstanding.
For the second patch, you'd better do that too imo.:)

BTW, the crash still happens as I told you before, we need to focus on it
because it's crash.

You couldn't really reproduce the crash any more?
Please 20 times next/prev on gst-play.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Michael Gratton <***@vee.net> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@vee.net

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #35 from Hyunjun Ko <***@igalia.com> ---
Regarding the problem #1, I proposed patches on the bug #796470, which is about
a bit of refactoring GstVaapiDisplay.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Hyunjun Ko <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371313|rejected |none
status| |
Attachment #371313|0 |1
is obsolete| |
Attachment #371315|reviewed |none
status| |
Attachment #371315|0 |1
is obsolete| |

--- Comment #36 from Hyunjun Ko <***@igalia.com> ---
Created attachment 372503
--> https://bugzilla.gnome.org/attachment.cgi?id=372503&action=edit
display: egl: fix to create VaapiDisplayEGL with native EGL display

gst_vaapi_display_egl_new_with_native_display has been broken since no
one uses it.

Currently it needs to call this api to create display with provided EGL
display so that it could avoid duplicated calls for the native display
(eg. eglTerminate).



-----------------------------------

I rebased on the patches on bug #796470.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Status|RESOLVED |REOPENED
Resolution|FIXED |---

--- Comment #38 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Reopen to fix this issue in master

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution|--- |FIXED
Target Milestone|git master |1.14.2

--- Comment #37 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
I have pushed these commit in branch 1.14

* fdf7f65c libs: display: resurrect parent private memember
* 55a14e39 libs: display: egl: initialize params structure
* fa54262a plugins: handle EGL when creating VAAPI display from gl
* 67d386d9 display: egl: fix to create VaapiDisplayEGL with native EGL display

@hyunjun commit 67d386d9 is a modified version of yours. Also, I suspect we
really need to resurrect the parent member in private structure (commit
fdf7f65c), because of the locks to the wrapped vaDisplay.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Depends on| |796470

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #39 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372637
--> https://bugzilla.gnome.org/attachment.cgi?id=372637&action=edit
display: egl: create VaapiDisplayEGL with native EGL display

gst_vaapi_display_egl_new_with_native_display() has been broken since
it wasn't used.

Currently it's needed to call this API to create a display providing
the EGL display, so it could avoid duplicated calls to the native
display (eg. eglTerminate).

Signed-off-by: Victor Jaquez <***@igalia.com>

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #371465|0 |1
is obsolete| |

--- Comment #40 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372638
--> https://bugzilla.gnome.org/attachment.cgi?id=372638&action=edit
plugins: handle EGL when creating VAAPI display from gl

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #41 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372639
--> https://bugzilla.gnome.org/attachment.cgi?id=372639&action=edit
libs: display: egl: initialize params structure

Statically initialise the internal params structure.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #42 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372640
--> https://bugzilla.gnome.org/attachment.cgi?id=372640&action=edit
libs: display: resurrect parent private member

This is, practically, a revert of commit dcf135e2.

The parent logic is useful for the EGL display, which is a decorator
of the real windowing subsystem (X11 or Wayland). Thus it is avoided
calling vaInitialize() and vaTerminate() twice.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372503|0 |1
is obsolete| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Blocks| |796564

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Blocks| |796470
Depends on|796470 |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372637|0 |1
is obsolete| |

--- Comment #43 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372653
--> https://bugzilla.gnome.org/attachment.cgi?id=372653&action=edit
display: egl: create VaapiDisplayEGL with native EGL display

gst_vaapi_display_egl_new_with_native_display() has been broken since
it wasn't used.

Currently it's needed to call this API to create a display providing
the EGL display, so it could avoid duplicated calls to the native
display (eg. eglTerminate).

Signed-off-by: Victor Jaquez <***@igalia.com>

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372638|0 |1
is obsolete| |

--- Comment #44 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372654
--> https://bugzilla.gnome.org/attachment.cgi?id=372654&action=edit
plugins: handle EGL when creating VAAPI display from gl

If GstGL reports a EGL platform force to create a EGL display using
the native EGL display.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372639|0 |1
is obsolete| |

--- Comment #45 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372655
--> https://bugzilla.gnome.org/attachment.cgi?id=372655&action=edit
libs: display: egl: initialize params structure

Statically initialise the internal params structure.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372640|0 |1
is obsolete| |

--- Comment #46 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Created attachment 372656
--> https://bugzilla.gnome.org/attachment.cgi?id=372656&action=edit
libs: display: resurrect parent private member

This is, practically, a revert of commit dcf135e2.

The parent logic is useful for the EGL display, which is a decorator
of the real windowing subsystem (X11 or Wayland). Thus it is avoided
calling vaInitialize() and vaTerminate() twice.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372656|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution|--- |FIXED

--- Comment #47 from Víctor Manuel Jáquez Leal <***@igalia.com> ---
Attachment 372653 pushed as 50470ae - display: egl: create VaapiDisplayEGL with
native EGL display
Attachment 372654 pushed as e62f3d7 - plugins: handle EGL when creating VAAPI
display from gl
Attachment 372655 pushed as ad974b4 - libs: display: egl: initialize params
structure
Attachment 372656 pushed as b0bebeb - libs: display: resurrect parent private
member

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372654|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372653|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Attachment #372655|none |committed
status| |

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

Víctor Manuel Jáquez Leal <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
Target Milestone|1.14.2 |1.15.1

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=795391

--- Comment #48 from Hyunjun Ko <***@igalia.com> ---
I didn't focus on this issue enough recently, sorry.
And thank you for your work with much better patches.