BUG: Timelapse rollover at midnight does not always happen

[13:53] <moonaddict> I am (happily!) using motion 3.1.17. the timelapse mpegs sometimes do not rotate at midnight  each day, but keep going in the same file for up to 72 hours... any ideas?
[13:55] <Lavr> How have you setup the filename?
[13:59] <moonaddict> wait
[14:00] <moonaddict> timelapse_filename %Y%m%d-timelapse
[14:00] <moonaddict> and
[14:00] <moonaddict> ffmpeg_timelapse_mode daily
[14:02] <Lavr> Hmm. That should work. Maybe if the machine is very busy at midnight Motion misses the event.
[14:02] <Lavr> If this is the case it is a real bug in Motion
[14:03] <moonaddict> Lavr: the busy thing could be a problem as cron & co do lots of stuff right then.
[14:04] <moonaddict> still - it happens sporadically
[14:04] <moonaddict> it's no big deal though
[14:05] <moonaddict> by the way: when monitoring axis 205 cameras, they get stuck after a few hours/days. so in my setup, I have scripted to reboot them regularily, so everything keeps working. as I've seen that with 6 205 cams so far, could be worth mentioning in the wiki somewhere...
[14:08] <Lavr> I will release  3.1.18_snap11 within the hour. It may have resolved all of this. Right Sack? :-)
[14:09] <moonaddict> woooo-hoo :)
[14:11] <Lavr> What framerate do you use?
[14:11] <Lavr> And do you run lowcpu?
[14:15] <moonaddict> framerate 2
[14:15] <moonaddict> low_cpu 0
[14:16] <moonaddict> but on another machine, it's low_cpu 1 -> same behaviour
[14:16] <Lavr> I can see in the code that if Motion misses a few frames because of high load from for example a cron job, it will wait till the next day. I need to make this more robust.
[14:17] <Lavr> if (cnt->currenttime->tm_min == 0 && cnt->currenttime->tm_sec == 0 && cnt->shots == 0) {
[14:18] <Lavr> If the code misses the single frame where this is valid it misses the timelapse reset
[14:19] <Lavr> I need to make sure it only gets valid ONCE and in a way so this long IF does not have to run all for each frame captured.
[14:20] <Lavr> But on the other hand make sure that if the event was missed it will catch it a few seconds later
[14:21] <moonaddict> hmm
[14:21] <Lavr> bbl. I will think about a fix for this. Maybe it will be in the snap11. it should not be too difficult to resolve.
[14:21] <moonaddict> that explains it.
[14:21] <moonaddict> :)

Test case

Environment

Follow up

Fix record