return0e has quit [Remote host closed the connection]
return0e has joined #linux-amlogic
return0e has quit [Remote host closed the connection]
sputnik_ has joined #linux-amlogic
jakogut has quit [Remote host closed the connection]
jakogut has joined #linux-amlogic
<ndufresne>
mjourdan, just given a shot of the tip of your branch, adaptive streaming in GStreamer moved from sometime working to hang or crash (including crash in vdec_isr+0x14/0x30)
_whitelogger has joined #linux-amlogic
Barada has joined #linux-amlogic
ldevulder_vacati has quit [Ping timeout: 272 seconds]
The_CooIest has quit [Ping timeout: 245 seconds]
The_Coolest has joined #linux-amlogic
The_CooIest has joined #linux-amlogic
The_CooIest has quit [Quit: FOLD For The Cause!!!]
indy has quit [Read error: Connection reset by peer]
indy has joined #linux-amlogic
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
return0e has joined #linux-amlogic
Darkmatter66_ has quit [Ping timeout: 246 seconds]
AntonioND has joined #linux-amlogic
_whitelogger has joined #linux-amlogic
<mjourdan>
ndufresne: do you have a command line so I can reproduce ?
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
Darkmatter66 has joined #linux-amlogic
_whitelogger has joined #linux-amlogic
_whitelogger has quit [Excess Flood]
_whitelogger has joined #linux-amlogic
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
afaerber has quit [Quit: Leaving]
afaerber has joined #linux-amlogic
afaerber has quit [Quit: Leaving]
Darkmatter66 has quit [Max SendQ exceeded]
Darkmatter66 has joined #linux-amlogic
afaerber has joined #linux-amlogic
sputnik__ has joined #linux-amlogic
sputnik_ has quit [Ping timeout: 252 seconds]
<ndufresne>
mjourdan, H264 mostly, H265 hangs on resets now, it used to work
<ndufresne>
I guess you could also give it a try with vlc/mpv, etc.
Barada has quit [Quit: Barada]
sputnik__ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
sputnik_ has quit [Remote host closed the connection]
sputnik_ has joined #linux-amlogic
<mjourdan>
ndufresne: alright thanks, hopefully I can reproduce
sputnik_ has quit [Read error: Connection reset by peer]
sputnik__ has joined #linux-amlogic
<ndufresne>
mjourdan, you could also use gst-play-1.0 on normal files, and use right/left arrow, which also hangs or crash
<ndufresne>
it happens when you quickly streamoff, requeue, streamon
<ndufresne>
though in mpd case, it's not clear, since it does not get very far
<ndufresne>
maybe it does CMD_STOP, streamoff/on and continue
sputnik__ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
<mjourdan>
ndufresne: kk. Once bitbake has finished recompiling stuff I'll see to fixing it, I really hope it's an easy one :D .
Lyude has quit [Quit: WeeChat 2.2]
Lyude has joined #linux-amlogic
sputnik__ has joined #linux-amlogic
sputnik_ has quit [Ping timeout: 244 seconds]
AntonioND has quit [Quit: :wq]
sputnik__ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
jelly has quit [Read error: Connection reset by peer]
jelly-home has joined #linux-amlogic
sputnik_ has quit [Remote host closed the connection]
sputnik_ has joined #linux-amlogic
jelly-home is now known as jelly
vagrantc has joined #linux-amlogic
rubdos has joined #linux-amlogic
random_yanek has quit [Quit: random_yanek]
random_yanek has joined #linux-amlogic
<mjourdan>
ndufresne: (on 1.14.2) something weird I'm seeing is that when I seek, after streamoff/streamon, the new pipeline is configured with 31 capture buffers. Is that on purpose ?
<mjourdan>
which shouldn't even be possible since the max buffers is 24.. mmh..
<mjourdan>
nevermind what I said, it's a bug on my side although I don't know where. For some reasons v4l2 tells me we have 31 buffers available after the streamoff/streamon sequence
<mjourdan>
ndufresne: it's so weird.. in between streamoff and streamon there are 31 capture buffers that get queued, but they come out of nowhere.
<ndufresne>
mjourdan, seems wrong
<ndufresne>
31 ...
<ndufresne>
some code is confused, seems like a good case to investigate
<ndufresne>
you can send me the GST_DEBUG="v4l2*:7" gst-play-1.0 ... 2> gst.log
<ndufresne>
I might be able to shed some light
<ndufresne>
otherwise I can try and produce it myself tonight
<mjourdan>
one index is skipped so it's indeed surprisingly 31. It's a state that shouldn't happen in the v4l2 core but I get no warning or anything
<ndufresne>
mjourdan, yesterday I was looking at how to enable some traces in your driver, do you have a hint ?
<mjourdan>
I'll run it with GST_DEBUG to see, but I have a feeling the bug is in the kernel
<mjourdan>
mmh nope, seems like gst keeps the first 8 buffers and then decides to add 23-24 more lol. https://dpaste.de/CLRz
<mjourdan>
Weirdness starts happening at line ~2192
<mjourdan>
another weird thing: buffer 4 just disappears and is not part of the "A buffer was lost, reallocating it" iteration
<mjourdan>
(oh, and I had to semi-revert the minimum amount of buffers, set it to 8.. Otherwise decode would get stuck with my sample with the default 4)
<ndufresne>
" allocated 8 mmap buffers out of 4 requested", I thought you had implement the associated CID
<ndufresne>
mjourdan, "Uncertain or not enough buffers, enabling copy threshold", are you using fakesink ?
<ndufresne>
the buffers are copied at decoder output in your pipeline
<ndufresne>
mjourdan, run valgrind on that thing, you got memory corruption
AntonioND has joined #linux-amlogic
<ndufresne>
if not, I don't know what is happending
<ndufresne>
mjourdan, do yocto patches gst-plugins-good ?
* ndufresne
don't trust any patches in yocto in general
<ndufresne>
mjourdan, actually, if this is not a corruption, it means that CREATE_BUFS is being called, which is odd
<ndufresne>
do you implement that ?
<ndufresne>
I never seen a decoder that implements that
<ndufresne>
32 being the maximum v4l2 can give
<ndufresne>
mjourdan, I'll try to reproduce, I don't know why it enters this callback loop
<ndufresne>
you should not implement CREATE_BUFS if you don't support it
<ndufresne>
ok, I think I see
<ndufresne>
the code didn't expect the driver to implement CREATE_BUFS
<ndufresne>
that I will fix
sputnik_ has quit [Read error: Connection reset by peer]
sputnik_ has joined #linux-amlogic
afaerber has quit [Quit: Leaving]
afaerber has joined #linux-amlogic
AntonioND has quit [Quit: Quit]
Darkmatter66_ has joined #linux-amlogic
Darkmatter66 has quit [Ping timeout: 252 seconds]
return0e has quit [Remote host closed the connection]