crash at 1:30am almost every night

[johnsonx42]

well now I'm getting these "async stop taking too long" crashes way more often. I'd stopped getting these almost completely, and now 3 in a week (one earlier today for some reason didn't produce a crash.mdmp). it crashed at 12:30am trying to end the recording of "King of the Hill". The recording itself is fine, it ended on time at 12:30am and is playable.

logs attached, in case anything leads you to a possible fix.

[sub]

I checked this dump, and the crash was the same as the last one I posted the screenshot for. ie, inside MSNP.AX in the UpdateSignalAndServiceStatus function.

I'm guessing this is a function that is called regularly behind the scenes in MSNP.ax on some background thread, but it has some logic flaw that means its not entirely thread safe, and occasionally when the graph stops mid way through one of those UpdateSignalAndServiceStatus() calls, it crashes when the graph component it was about to use disappears (or is stopped or similar).

I'm not really sure what I could do to work around this. There might be some random combinations of steps I could perform that might help, but it's hard to guess what they might be.

I could try giving you a patch to force it to use the old 'Microsoft ATSC Network Provider' MSDvbNP.ax instead of the newer 'Microsoft Network Provider' MSNP.ax. It might help if this older version doesnt have the same problem. Alternatively, it might behave exactly the same if one just delegates its functionality to the other.

Or we could look at an option to never stop the graph. ie, it's always running, but not be capturing any files between recordings.

[sub]

You could see if this helps. It attempts to explicitly stop the network provider, waits 200 ms, then stops the main graph. It also adds a couple of logging message around these new activities.

[johnsonx42]

thanks sub, I'll put that patch in today. unfortunately the nature of this crash type is that it might be weeks before it comes up again, or it might come up again tomorrow.

it does seem that since your second patch to NPVRTSMon.ax which added defensive checks, I've not had any occurance of the original 1:30am crash in the exact same scenario. the crash I reported on 7-24 did look like the same crash, though it happened at 1:35am instead because the recording sequence was a little different; but that's the one where we were out of sync on NPVRTSMon versions so the crash dump was no help.

so it does look like we're making progress. fingers crossed!

[johnsonx42]

I guess I typed too loud... the 1:30am crash has returned tonight, actually at 1:31am (timestamp of the crash.mdmp). The suspect Zula Patrol recording is 1 minute, 24 seconds long. What's interesting about this crash is that part of NRecord survived long enough to cleanly end the Grey's Anatomy recording at 1:35am, and it even marked it successful in the database.

[sub]

That one was in npvrtsmon.ax, but unfortunately no longer matches my version, so I cant tell exactly where the crash occurred.

[johnsonx42]

argh!

[johnsonx42]

can you send me your version of NPVRTSMon.ax when you get it a little stabilized? Or have you moved beyond 2.1.5 compatibility now?

[johnsonx42]

You could see if this helps. It attempts to explicitly stop the network provider, waits 200 ms, then stops the main graph. It also adds a couple of logging message around these new activities.

I've had two of these 'graph won't stop' crashes in two days. No crash.mdmp created yesterday, but I did get one tonight (it didn't pop up the message saying it made a crash.mdmp tonight, but when I checked later I found it was there). Last night also I had to reboot the machine to get NRecord going again, but tonight it did exit cleanly.

The Family Guy recording tonight that it crashed trying to stop did stop on time and is playable, but is of course marked "Failed"

Yesterday's crash was also on Family Guy; that recording is zero length. there was another recording going on at the same time, Frasier on a different channel and tuner which completed without issue; that recording was fine as far as I know (it's been deleted, but my 'processing.log' indicates comskip ran on Frasier for the full 30 minutes, while it aborted in less than 5 seconds on the Family Guy recording)

I don't know if any of the log messages will help.

final edit: ok, I'm done changing this message. the above is now the correct story.

[johnsonx42]

sub, sorry I keep changing the post above, I keep realizing I'm not giving you the correct story... please wait a moment

ok, the version above is my final answer. thanks for your patience, if any