[MPlayer-users] MPEG/DVD crashing when taking screenshot
Josh
joshf87 at live.com
Sun Feb 14 23:25:18 CET 2016
On 2/14/2016 15:37, Roberto Togni wrote:
> On Tue, 20 Oct 2015 04:36:37 -0400
> Josh <joshf87 at live.com> wrote:
>
>> On 10/13/2015 16:00, Roberto Togni wrote:
>>> On Fri, 25 Sep 2015 23:23:45 +0200
>>> Roberto Togni <rxt at rtogni.it> wrote:
>>>
>>>> On Thu, 24 Sep 2015 00:04:37 +0200
>>>> Roberto Togni <rxt at rtogni.it> wrote:
>>>>
>>>>> On Tue, 22 Sep 2015 00:38:28 -0400
>>>>> Josh <joshf87 at live.com> wrote:
>>>>>
>>>>>> On 9/21/2015 16:42, Roberto Togni wrote:
>>>>>>> On Mon, 21 Sep 2015 01:05:59 -0400
>>>>>>> Josh <joshf87 at live.com> wrote:
>>>>>>>
>>>>>>>> On 9/14/2015 17:40, Roberto Togni wrote:
>>>>>>>>> On Mon, 14 Sep 2015 20:59:46 +0200
>>>>>>>>> Roberto Togni <rxt at rtogni.it> wrote:
>>>>>>>>>
>>>>>>>>>> On Sun, 13 Sep 2015 00:26:42 -0400
>>>>>>>>>> Josh <joshf87 at live.com> wrote:
>>>>>>>>>>
>>>>>>>>>>> I compiled r37524 with a newer FFmpeg and got different but still
>>>>>>>>>>> seemingly useless results from gdb so I ran Dr Memory, both logs are
>>>>>>>>>>> included in the tar.xz. The logs themselves will get me queued for
>>>>>>>>>>> moderation for being too big:
>>>>>>>>>>>
>>>>>>>>>>> http://sourceforge.net/projects/mplayerwin/files/Misc/mpl-bin-r37524-debug-2.tar.xz/download
>>>>>>>>>>
>>>>>>>>>> Hi,
>>>>>>>>>> just to keep you updated, I was able to reproduce it with your sample
>>>>>>>>>> and get a valid backtrace on linux.
>>>>>>>>>>
>>>>>>>>>> ...
>>>>>>>>>> sending VFCTRL_SCREENSHOT!
>>>>>>>>>>
>>>>>>>>>> Program received signal SIGSEGV, Segmentation fault.
>>>>>>>>>> 0x0000555555f68625 in yuv2rgb_X_c_template (hasAlpha=0,
>>>>>>>>>> target=AV_PIX_FMT_RGB24, y=13, dstW=2386, dest=0x555558be4680 "",
>>>>>>>>>> alpSrc=0x0, chrFilterSize=4, chrVSrc=0x5555572d9a48,
>>>>>>>>>> chrUSrc=0x55555711f128, chrFilter=0x5555572d76c8, lumFilterSize=1,
>>>>>>>>>> lumSrc=0x55555712c5e8, lumFilter=0x55555711e8ba, c=<optimized out>)
>>>>>>>>>> at libswscale/output.c:1340
>>>>>>>>>> 1340 Y1 += lumSrc[j][i * 2] * lumFilter[j];
>>>>>>>>>> (gdb) bt
>>>>>>>>>> #0 0x0000555555f68625 in yuv2rgb_X_c_template (hasAlpha=0,
>>>>>>>>>> target=AV_PIX_FMT_RGB24, y=13, dstW=2386, dest=0x555558be4680 "",
>>>>>>>>>> alpSrc=0x0, chrFilterSize=4, chrVSrc=0x5555572d9a48,
>>>>>>>>>> chrUSrc=0x55555711f128, chrFilter=0x5555572d76c8, lumFilterSize=1,
>>>>>>>>>> lumSrc=0x55555712c5e8, lumFilter=0x55555711e8ba, c=<optimized out>)
>>>>>>>>>> at libswscale/output.c:1340
>>>>>>>>>> #1 yuv2rgb24_X_c (c=0x555557223520, lumFilter=0x55555711e8ba,
>>>>>>>>>> lumSrc=0x55555712c5e8, lumFilterSize=1, chrFilter=0x5555572d76c8,
>>>>>>>>>> chrUSrc=0x55555711f128, chrVSrc=0x5555572d9a48, chrFilterSize=4,
>>>>>>>>>> alpSrc=0x0, dest=0x555558be4680 "", dstW=2386, y=13)
>>>>>>>>>> at libswscale/output.c:1515
>>>>>>>>>> #2 0x0000555555f4dc9e in packed_vscale (c=<optimized out>,
>>>>>>>>>> desc=<optimized out>, sliceY=13, sliceH=<optimized out>)
>>>>>>>>>> at libswscale/vscale.c:129
>>>>>>>>>> #3 0x0000555555f3dc81 in swscale (c=0x5555572b12c0,
>>>>>>>>>> src=0x55555711e8a0, srcStride=0x1000, srcSliceY=1460470520,
>>>>>>>>>> srcSliceH=1460470480, dst=0x55555711f128, dstStride=0x7fffffffca80) at
>>>>>>>>>> libswscale/swscale.c:682 #4 0x0000555555f3ef4b in sws_scale
>>>>>>>>>> (c=0x5555572b12c0, srcSlice=0x7fffffffca90, srcStride=0x55555712c5e8,
>>>>>>>>>> srcSliceY=1, srcSliceH=1462597320, dst=0x7fffffffcab0,
>>>>>>>>>> dstStride=0x5555570f4c60) at libswscale/swscale.c:1263
>>>>>>>>>> #5 0x00005555558295da in draw_slice (vf=0x5555570f53b0,
>>>>>>>>>> src=0x7fffffffcb90, stride=0x7fffffffcb80, w=1920, h=16, x=1460791456,
>>>>>>>>>> y=16)
>>>>>>>>>> at libmpcodecs/vf_screenshot.c:64
>>>>>>>>>> #6 0x0000555555826021 in draw_slice (s=<optimized out>, src=<optimized
>>>>>>>>>> out>, offset=<optimized out>, y=<optimized out>, type=<optimized out>,
>>>>>>>>>> height=<optimized out>) at libmpcodecs/vd_ffmpeg.c:588
>>>>>>>>>> #7 0x0000555555eb48e4 in ff_draw_horiz_band (avctx=0x5555570f69a0,
>>>>>>>>>> cur=<optimized out>, last=<optimized out>, y=16, h=16,
>>>>>>>>>> picture_structure=3, first_field=0, low_delay=0)
>>>>>>>>>> at libavcodec/mpegutils.c:77
>>>>>>>>>> #8 0x0000555555b95135 in ff_mpeg_draw_horiz_band
>>>>>>>>>> (s=s at entry=0x5555570f6e40, y=<optimized out>, h=<optimized out>) at
>>>>>>>>>> libavcodec/mpegvideo.c:2720 #9 0x0000555555b60db5 in mpeg_decode_slice
>>>>>>>>>> (s=s at entry=0x5555570f6e40, mb_y=mb_y at entry=1,
>>>>>>>>>> buf=buf at entry=0x7fffffffcdd8, buf_size=buf_size at entry=2077) at
>>>>>>>>>> libavcodec/mpeg12dec.c:1927 #10 0x0000555555b63b35 in decode_chunks
>>>>>>>>>> (avctx=avctx at entry=0x5555570f69a0,
>>>>>>>>>> picture=picture at entry=0x5555570f6760,
>>>>>>>>>> got_output=got_output at entry=0x7fffffffcf60, buf=0x555556f088c0 "",
>>>>>>>>>> buf_size=2111) at libavcodec/mpeg12dec.c:2759 #11 0x0000555555b645ab in
>>>>>>>>>> mpeg_decode_frame (avctx=0x5555570f69a0, data=0x5555570f6760,
>>>>>>>>>> got_output=0x7fffffffcf60, avpkt=<optimized out>) at
>>>>>>>>>> libavcodec/mpeg12dec.c:2836 #12 0x0000555555c6762e in
>>>>>>>>>> avcodec_decode_video2 (avctx=0x5555570f69a0, picture=0x5555570f6760,
>>>>>>>>>> got_picture_ptr=0x7fffffffcf60, avpkt=0x7fffffffcf90) at
>>>>>>>>>> libavcodec/utils.c:2096 ---Type <return> to continue, or q <return> to
>>>>>>>>>> quit--- #13 0x0000555555826e5f in decode (sh=0x555556edc410,
>>>>>>>>>> data=0x1000, len=1460848104, flags=-12400) at
>>>>>>>>>> libmpcodecs/vd_ffmpeg.c:949 #14 0x0000555555775f4f in decode_video
>>>>>>>>>> (sh_video=0x555556edc410, start=0x555556f088c0 "", in_size=2111,
>>>>>>>>>> drop_frame=0, pts=1.083922266960144, full_frame=0x7fffffffd0b0)
>>>>>>>>>> at libmpcodecs/dec_video.c:398
>>>>>>>>>> #15 0x00005555556dd772 in update_video (blit_frame=0x7fffffffd158)
>>>>>>>>>> at mplayer.c:2484
>>>>>>>>>> #16 0x00005555556d0980 in main (argc=5, argv=0x7fffffffe288) at
>>>>>>>>>> mplayer.c:3811
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> You're probably right with the commit that broke it, I'm checking if
>>>>>>>>>> swscale needs to be called differently now.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> May be a problem in ffmpeg, see https://trac.ffmpeg.org/ticket/4850
>>>>>>>>>
>>>>>>>>> I can get a crash with similar backtrace with
>>>>>>>>> ffmpeg -i mpeg_crash_screenshot.mpg -dsth 1096 -c png -f null /dev/null
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Program received signal SIGSEGV, Segmentation fault.
>>>>>>>>> 0x0000000000de5a57 in yuv2rgb_write (hasAlpha=0,
>>>>>>>>> target=AV_PIX_FMT_RGB24,
>>>>>>>>> y=1088, _b=0x1e0f266, _g=0x1e0f266, _r=0x1e0f266, A2=0, A1=0,
>>>>>>>>> Y2=<optimized out>, Y1=<optimized out>, i=<optimized out>,
>>>>>>>>> _dest=0x7fffcc7f8040 "") at libswscale/output.c:1246
>>>>>>>>> 1246 dest[i * 6 + 0] = r_b[Y1];
>>>>>>>>> (gdb) bt
>>>>>>>>> #0 0x0000000000de5a57 in yuv2rgb_write (hasAlpha=0,
>>>>>>>>> target=AV_PIX_FMT_RGB24,
>>>>>>>>> y=1088, _b=0x1e0f266, _g=0x1e0f266, _r=0x1e0f266, A2=0, A1=0,
>>>>>>>>> Y2=<optimized out>, Y1=<optimized out>, i=<optimized out>,
>>>>>>>>> _dest=0x7fffcc7f8040 "") at libswscale/output.c:1246
>>>>>>>>> #1 yuv2rgb_X_c_template (hasAlpha=0, target=AV_PIX_FMT_RGB24, y=1088,
>>>>>>>>> dstW=1920, dest=0x7fffcc7f8040 "", alpSrc=0x0, chrFilterSize=4,
>>>>>>>>> chrVSrc=0x1b4dd30, chrUSrc=0x1b44c70, chrFilter=0x1b6e0c0,
>>>>>>>>> lumFilterSize=4, lumSrc=0x1b4e4f8, lumFilter=0x1b67440,
>>>>>>>>> c=<optimized out>)
>>>>>>>>> at libswscale/output.c:1370
>>>>>>>>> #2 yuv2rgb24_X_c (c=0x1ba75c0, lumFilter=0x1b67440, lumSrc=0x1b4e4f8,
>>>>>>>>> lumFilterSize=4, chrFilter=0x1b6e0c0, chrUSrc=0x1b44c70,
>>>>>>>>> chrVSrc=0x1b4dd30, chrFilterSize=4, alpSrc=0x0, dest=0x7fffcc7f8040
>>>>>>>>> "",
>>>>>>>>> dstW=1920, y=1088) at libswscale/output.c:1515
>>>>>>>>> #3 0x0000000000dca58e in packed_vscale (c=<optimized out>,
>>>>>>>>> desc=<optimized out>, sliceY=1088, sliceH=<optimized out>)
>>>>>>>>> at libswscale/vscale.c:129
>>>>>>>>> #4 0x0000000000dc5d31 in swscale (c=0x1b4e980, src=0x1e0f266,
>>>>>>>>> srcStride=0x0,
>>>>>>>>> srcSliceY=28632888, srcSliceH=28632848, dst=0x1b44c70,
>>>>>>>>> dstStride=0x7fffffffd6f0) at libswscale/swscale.c:682
>>>>>>>>> #5 0x0000000000dc70ba in sws_scale (c=0x1b4e980,
>>>>>>>>> srcSlice=0x7fffffffd700,
>>>>>>>>> srcStride=0x1b4e4f8, srcSliceY=4, srcSliceH=28762304,
>>>>>>>>> dst=0x7fffffffd720,
>>>>>>>>> dstStride=0x7fffffffd7d0) at libswscale/swscale.c:1263
>>>>>>>>> #6 0x000000000050a088 in scale_slice (y=<optimized out>,
>>>>>>>>> link=<optimized out>, field=<optimized out>, mul=<optimized out>,
>>>>>>>>> h=<optimized out>, sws=<optimized out>, cur_pic=<optimized out>,
>>>>>>>>> out_buf=<optimized out>) at libavfilter/vf_scale.c:477
>>>>>>>>> #7 filter_frame (link=0x1b44520, in=0x1e0ecc0) at
>>>>>>>>> libavfilter/vf_scale.c:579
>>>>>>>>> #8 0x000000000049f1df in ff_filter_frame_framed (link=0x1b44520,
>>>>>>>>> frame=0x1e0ecc0) at libavfilter/avfilter.c:1089
>>>>>>>>> #9 0x00000000004a1299 in ff_filter_frame (link=0x1b44520,
>>>>>>>>> frame=0x1e0ecc0)
>>>>>>>>> at libavfilter/avfilter.c:1173
>>>>>>>>> #10 0x000000000049f1df in ff_filter_frame_framed (link=0x1b44b20,
>>>>>>>>> frame=0x1e0ecc0) at libavfilter/avfilter.c:1089
>>>>>>>>> #11 0x00000000004a1299 in ff_filter_frame (link=0x1b44b20,
>>>>>>>>> frame=0x1e0ecc0)
>>>>>>>>> at libavfilter/avfilter.c:1173
>>>>>>>>> #12 0x00000000004a53a2 in request_frame (link=0x1b44b20)
>>>>>>>>> at libavfilter/buffersrc.c:378
>>>>>>>>> #13 0x00000000004a50d4 in av_buffersrc_add_frame_internal (
>>>>>>>>> ctx=ctx at entry=0x1b455e0, frame=frame at entry=0x1cca620,
>>>>>>>>> flags=flags at entry=4)
>>>>>>>>> at libavfilter/buffersrc.c:180
>>>>>>>>> #14 0x00000000004a55cd in av_buffersrc_add_frame_flags (ctx=0x1b455e0,
>>>>>>>>> frame=frame at entry=0x1cca620, flags=flags at entry=4)
>>>>>>>>> at libavfilter/buffersrc.c:105
>>>>>>>>> #15 0x0000000000487272 in decode_video (ist=0x1b3ca00, pkt=<optimized
>>>>>>>>> out>,
>>>>>>>>> got_output=<optimized out>) at ffmpeg.c:2179
>>>>>>>>> #16 0x00000000004715d3 in process_input_packet (pkt=<optimized out>,
>>>>>>>>> ist=<optimized out>) at ffmpeg.c:2327
>>>>>>>>> #17 process_input (file_index=<optimized out>) at ffmpeg.c:3833
>>>>>>>>> #18 transcode_step () at ffmpeg.c:3921
>>>>>>>>> #19 transcode () at ffmpeg.c:3974
>>>>>>>>> #20 main (argc=28559872, argv=0x7fffffffe010) at ffmpeg.c:4157
>>>>>>>>> (gdb)
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Seems that the problem is related to slices, in MPlayer it works for the
>>>>>>>>> first slice but fails for the 2nd (slice height is 16; offset 0 is ok,
>>>>>>>>> offset 16 crashes).
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> The attached workaroud fixes the crash, but produces a wrong picture;
>>>>>>>>> unless I missed some stupid detail, it should be equivalent to the old
>>>>>>>>> code.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Ciao,
>>>>>>>>> Roberto
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> MPlayer-users mailing list
>>>>>>>>> MPlayer-users at mplayerhq.hu
>>>>>>>>> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
>>>>>>>>>
>>>>>>>>
>>>>>>>> I compiled the latest MPlayer trunk/FFmpeg master that has a fix for
>>>>>>>> FFmpeg #4850 and without your patch.
>>>>>>>>
>>>>>>>> There is no crash but it simply quits (no screenshot saved) after
>>>>>>>> 'sending VFCTRL_SCREENSHOT!' when slices are enabled, with the exception
>>>>>>>> of mpeg_crash_ss_beetlejuice.mpg it quits slices or noslices.
>>>>>>> Fix for #4850 did not stop the crash for me too.
>>>>>>>
>>>>>>> What I found out is that if I force FFmpeg to use the old scaler
>>>>>>> filters, everything works fine.
>>>>>>>
>>>>>>> I was still unable to find out if something changed in the way you need
>>>>>>> to use swscaler (I don't think so), if there is a bug in the new
>>>>>>> swscaler, or if we always used it in the wrong way and we were just
>>>>>>> lucky up to now.
>>>>>>>
>>>>>>> I'm not much worried about vf_screenshot, but about all other uses of
>>>>>>> the scaler code in the mail player.
>>>>>>>
>>>>>>> If you want to test it, just comment out #define NEW_FILTER at the end
>>>>>>> of ffmpeg/libswscale/swscale_internal.h and recompile everything.
>>>>>>>
>>>>>>> This is a possible workaround for people using the internal ffmpeg copy
>>>>>>> of MPlayer, it will not help if you use a system-wide install of ffmpeg.
>>>>>>>
>>>>>>>
>>>>>>> Ciao,
>>>>>>> Roberto
>>>>>>> _______________________________________________
>>>>>>> MPlayer-users mailing list
>>>>>>> MPlayer-users at mplayerhq.hu
>>>>>>> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
>>>>>>>
>>>>>>
>>>>>> As of current FFmpeg master with disabling the new filter everything
>>>>>> works fine so far.
>>>>> I found a way to produce a simple test case that fails in the exact way
>>>>> as MPlayer.
>>>>> See ffmpeg ticket https://trac.ffmpeg.org/ticket/4877
>>>>>
>>>> Crash is fixed with slices with latest ffmpeg, but image is wrong.
>>>> See https://trac.ffmpeg.org/ticket/4881
>>>>
>>>> The second sample still crashes with -noslices, but up to now I was
>>>> unable to create a simple test case to report it.
>>>>
>>> Fixed in FFmpeg git.
>>>
>>> The only problem remaining is the second sample with -noslices, it
>>> still crashes for me on Linux X86_64. I'll check if I can debug it
>>> in MPlayer, since I can't create a stand-alone testcase.
>>>
>>>
>>> Ciao,
>>> Roberto
>>> _______________________________________________
>>> MPlayer-users mailing list
>>> MPlayer-users at mplayerhq.hu
>>> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
>>>
>>
>> Much appreciation on fixing this issue so far. Thanks Roberto.
> Finally fixed.
> It was a MPlayer bug in the use of swscaler, not related to the new
> sws filter code. I have no idea why it was not crashing also with the
> old code.
>
> Ciao,
> Roberto
> _______________________________________________
> MPlayer-users mailing list
> MPlayer-users at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
>
I did a fresh checkout of FFmpeg from git a long time ago and forgot to
disable the new filter since then. 'beetlejuice' wasn't crashing anymore
slices/noslices but either way just produced a 0.99MB black screenshot
(it was the ONLY video doing that). r37733 actually fixed that.
I can't find any reason not to leave the new filter enabled.
More information about the MPlayer-users
mailing list