Trying to diagnose a READY --> PAUSED transition bug

[Nicolas Dufresne]

Le lun. 25 mars 2019 17 h 26, David Ing <ding at panopto.com> a écrit :

> I am using Gstreamer 1.14.4.  I am having an issue with one of my
> pipelines on Windows.  I attached an SVG file which shows my pipeline at
> the point in time when it gets stuck.  The issue occurs when I try to move
> my pipeline from READY to PAUSED.  I do this by calling
> gst_element_set_state(gstPipeline, GST_STATE_PAUSED), and after that call
> is made, I see all elements within the pipeline transition to PAUSED except
> the *GstFileSink which remains stuck in READY*.
>
> Also:
>
>    - The issue does not repro on Linux (only Windows).
>    - I use the same basic code to render many other kinds of jobs and I
>    never have a problem with the GstFileSink.  It is only for this one
>    specific pipeline that I have the problem.
>    - I am 100% certain that this isn't some kind of file system issue
>    (e.g. folder permissions).
>
> For logging I set the the "filesink" category to TRACE and this is the
> entirety of what it shows me:
>
> Pipeline is in NULL state, as I connect the muxer to the filesink I see
> this:
>
> 19-03-25T20:44:19.420327 INFO    filesink gstfilesink.c:294 filename :
> C:\dev\github\panopto\panopto-core\Debug\x64\test-scratch\2019-03-25T20-44-11.942351\CompositionTests\Debug\0ed40ba8-7d3d-4569-8294-aa02017efdfc-9531e8a6-d94f-44ed-a79a-aa0b0008a713.mp4
> 19-03-25T20:44:19.420327 INFO    filesink gstfilesink.c:295 uri      :
> file:///C:/dev/github/panopto/panopto-core/Debug/x64/test-scratch/2019-03-25T20-44-11.942351/CompositionTests/Debug/0ed40ba8-7d3d-4569-8294-aa02017efdfc-9531e8a6-d94f-44ed-a79a-aa0b0008a713.mp4
>
> Immediately after I call  gst_element_set_state(gstPipeline,
> GST_STATE_READY)  , I see:
>
> 19-03-25T20:44:19.458327 DEBUG   filesink gstfilesink.c:516 Seeking to
> offset 0 using fseeko
> 19-03-25T20:44:19.458327 DEBUG   filesink gstfilesink.c:416 opened file
> C:\dev\github\panopto\panopto-core\Debug\x64\test-scratch\2019-03-25T20-44-11.942351\CompositionTests\Debug\0ed40ba8-7d3d-4569-8294-aa02017efdfc-9531e8a6-d94f-44ed-a79a-aa0b0008a713.mp4,
> seekable 1
>
> 6.3 seconds after I call  gst_element_set_state(gstPipeline,
> GST_STATE_PAUSED)  , I see:
>
> 19-03-25T20:44:25.853308 DEBUG   filesink gstfilesink.c:586 Ignored
> SEGMENT event of format 3 (time)
>
> My best guess is that it's some kind of rare timing bug within gstreamer
> (which most pipelines are lucky to avoid).
>
> I am looking for tips about how to diagnose the exact cause of this bug:
> But keep in mind that I have no debug symbols because I'm on Windows.
>

Windows build have debug symbols, but only gdb format. You can download gdb
for Windows in order to trace. Best experience is if you have made your own
build, since then it will be able to map to the source code. Another note,
there has been load of work done to support a subset of the build, using
msvc, that is currently in GStreamer master of cerbero as a variant (-v
msvc iirc)

_______________________________________________
> gstreamer-devel mailing list
> gstreamer-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/gstreamer-devel/attachments/20190326/8a559953/attachment.html>