Also: #19974
Also: #19974
I think my HA is also affected by this. I haven’t done any firewall snooping and I’m not using pFsense, but I also have a timer based on Kodi’s ”turn_off” event, and the symptom of repeated ”turn_off” events match. The workaround from @ChopperRob above also seems to work.
(Probably unrelated, but I’ve also noticed that my ”turn display on when Kodi becomes idle” stops working when my LAN/WLAN gets very congested (such as when doing backups over the network). I’m using ”idle” because the log never shows a ”turn_on” event from Kodi; just ”idle”, ”playing” and the intermittently incessant ”turn_off” events.)
Seems to be fixed in VLC 3.0.9.2-1 (Ubuntu 20.04); at least I haven’t been able to reproduce it anymore.
Same ”double free or corruption (!prev)” reproducible here too (VLC 3.0.9.2-1, Ubuntu 20.04), even more easily: the position doesn’t matter, and the first attempt to screenshot is already enough to trigger the issue.
I came across this when trying to screenshot a Superman cartoon I downloaded from Internet Archive [1].
* [1] https://archive.org/download/dom-6746superman-episode7-theundergroundworld512kb/dom-6746superman-episode7-theundergroundworld512kb.mp4
I applied the patch on top of current head, built a deb, installed it (replacing the 4.4.0 release), and can confirm that the patch fixes the issue: the snapshots produced are now in the correct resolution (1920×1080) and look as they should (no glitching).
I’m running debsums -c daily, hoping to catch file corruption and/or other abnormalities early. (The Debian package even provides a pre-made cronjob for this task.)
debsums -c gitlab-ce reports .gitlab_kas_secret as deviating from the provided md5sum.
# debsums -c gitlab-ce
/opt/gitlab/embedded/service/gitlab-rails/.gitlab_kas_secret
# debsums -c gitlab-ce
#
ii gitlab-ce 14.5.1-ce.0 amd64 GitLab Community Edition (including NGINX, Postgres, Redis)
external_url 'http://localhost'
#2635 is basically the same, but I’ve never seen debsums complain about schema.rb.
In Motion 4.4.0, using recommended values of width and height results in broken snapshots, when netcam_high_url is used. Either the warning about using a different resolution is incorrect, or the creation of snapshots breaks when width and height are set so as to not cause a warning on startup.
I have a 1080p camera (a TP-Link Tapo C100) with a 360p substream, so I’ve specied both URLs as follows:
netcam_url "rtsp://192.168.1.206/stream2"
netcam_high_url "rtsp://192.168.1.206/stream1"
I’ve set up snapshots to occur with snapshot_interval = 60.
If I set width and height based on the main stream (1920×1080), Motion starts up with a warning:
netcam_rtsp_ntc: The network camera is sending pictures in a different
netcam_rtsp_ntc: size than specified in the configuration file.
netcam_rtsp_ntc: The picture is being transcoded into the size
netcam_rtsp_ntc: requested in the configuration. If possible change
netcam_rtsp_ntc: netcam or configuration to indicate the same size
netcam_rtsp_ntc: to possibly lower CPU usage.
netcam_rtsp_ntc: Netcam: 640 x 360 => Config: 1920 x 1080
This implies I should set width and height based on the lower resolution stream, so I’ve done that, and the warning goes away.
With Motion 4.3.2, the snapshots were saved as I would expect, albeit of course now in the lower resolution (360p), whereas I’d prefer them in the full 1080p.
After upgrading to 4.4.0, the snapshots are still 360p, but now their content is garbled and green:
I haven’t looked at the code, but what this looks like to me is as if Motion is now using the high resolution URL for the image data, but the lower resolution dimensions, and the resulting file is a (broken) crop of the full resolution image. Indeed, if I set width and height based on the full stream (i.e. back to 1920×1080), the snapshots now appear the full resolution and without corruption.
Pictures from detected motion (picture_output) do not appear to be affected, i.e. they are saved in the correct 1080p resolution and look fine, regardless of the values of width and height. Video output likewise appears in the higher resolution, as I would expect.
Here’s a minimal configuration that reproduces this issue for me, minus netcam_userpass.
log_level 8
target_dir "/home/jani/tmp"
netcam_url "rtsp://192.168.1.206/stream2"
netcam_high_url "rtsp://192.168.1.206/stream1"
netcam_params rtsp_transport=tcp
netcam_high_params rtsp_transport=tcp
width 640
height 360
snapshot_interval 10
Release v1.47.2 was mentioned in another issue’s comments, and Play Store version has indeed been updated, but that release is missing from Github releases.
Open Github releases page for the project.
Find release v.1.47.2 (or later).
The current latest release on Github is 1.47.0.
I don’t know if it’s related, but v.1.47.0 has been tagged ”changeme” instead of the usual vX.Y.Z pattern.
All right, I got curious enough try again, and found that this is apparently caused by the static build of ffmpeg that I’m using: I used yle-dl --verbose again, then interrupted the download immediately, ran the ffmpeg command (as reported by yle-dl) directly, and this resulted in a segmentation fault.
I then tried the same command in a VM with ffmpeg from Ubuntu repositories, and there the download finished successfully.
So the only remaining issue with yle-dl in this is the terminal being broken: that did not occur with directly-run ffmpeg despite the segfault.
I’m closing this issue though, so that it remains as documentation for the problem (and cause) as originally reported.
I tested 95.0a1 (Nightly), and opening the hamburger menu (using the steps I outlined above) works now.
Closing it is still partly broken though, as clicking on the menu button doesn’t close the menu; I think that is bug 1694514.
For a workaround, clicking outside the menu (and the menu button) does close the menu.