Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[bob]

This is the latest version of MC24 for Linux.  Please post bugs here.  Please start a new thread for anything requiring discussion.  Non-bug posts will be deleted.

Download:
amd64      http://files.jriver.com/mediacenter/channels/v24/latest/MediaCenter-24.0.28-amd64.deb
i386      http://files.jriver.com/mediacenter/channels/v24/latest/MediaCenter-24.0.28.deb
arm      http://files.jriver.com/mediacenter/channels/v24/latest/MediaCenter-24.0.28-armhf.deb

Also in the latest repo.

24.0.28 ( 5/14/2018)

1. NEW: Added info to the Reporter Advanced Information showing the system memory and free memory.
2. Changed: For Linux ARM, disable multithreaded drawfast code (used in resizing) until neon is fully implemented.
3. Changed: For Linux ARM, cap maximum threads and weight for thumbnailing until neon is fully implemented.
4. Fixed: Linux ALSA output device, wasn't configuring local zones other than the default zone properly.
5. Changed: Updated the gtk dependencies from libgtk2.0 to libgtk-3-0

24.0.26-2 (5/10/2018)

1. Changed: Switched the default skin to Modern Cards: Grey Edition.
2. NEW: Implemented support for direct DSD for devices that support DSD_U32_BE (i.e. iFi iDSD and other xmos based devices). See the device options under the Audio device config.

24.0.26 (5/10/2018)

1. Fixed: The file browser wasn't working in a few places (i.e. Languages->Browse Translations Folder)
2. Changed: Another tweak to registration. Will require re-entering the license key on intel based linux platforms.
3. NEW: Work on the integrated browser (not implemented yet) requires the addition of webkit2 library dependencies to the .deb
4. Fixed: Added the libgomp dependency to the .deb

24.0.19 (4/29/2018)

1. Fixed: Tooltip box shadows were offsetting the text below the shadow.
2. NEW: Text edit boxes now have a more convenient way to add Latin characters with diacritics.  Press and hold a letter key and a list of choices with diacritical versions for that letter is shown to choose from.
3. Fixed: On ARM the non-sse bilinear rendering wasn't properly handling the alpha channel resulting in some see through icons.

24.0.18 (4/24/2018)

1. Changed: Updated FFmpeg to version 4.0, used for file analysis, transcoding and video playback.
2. Changed: Optimized Video rendering image uploading for better performance and reliability.

24.0.16 (4/17/2018)

1. FIXED: A significant memory leak.

24.0.15 (4/13/2018)

1. NEW: Added support for bilinear image resizing (on ARM which didn't have it before) so that images in the program will look much better.
2. Fixed: A segfault upon startup with systems that have non-standard /dev tree.

24.0.12 (4/10/2018)

1. Fixed: Non-Root users on linux couldn't send WOL packets.
2. Fixed: Removed unsupported print commands from menus.
3. NEW: Much faster FFT code for places where spectrum analysis is used.
4. Changed: Some licensing code. Registration code will need to be re-entered.

24.0.2 (02/03/2018)

1. NEW: First MC24 build for linux

[mwillems]

What is neon?  How far along is implementation?

[Hendrik]

NEON is to ARM what SSE is to x86, ie. SIMD multimedia instructions that are well suited for image processing among other things.

[Awesome Donkey]

The output device issue is indeed fixed, thanks Bob. Also all my hw: devices show up and work fine with this build. Tested both PCM and DSD with both of my DACs/zones, and they're fine. Can't test native DSD output, since my Loki seems to require/use DoP. Oh well, it works nonetheless and the end result is the same.

Looks like smooth sailing.

[Mike Noe]

Huge mem leak of some kind:

Code: [Select]

The process mediacenter24 (with pid 1966) is using approximately 2.4 GB of memory.
It is using 2.4 GB privately, and a further 12.7 MB that is, or could be, shared with other programs.
Dividing up the shared memory between all the processes sharing that memory we get a reduced shared memory usage of 2.9 MB. Adding that to the private usage, we get the above mentioned total memory footprint of 2.4 GB.
Did not see this in 26.  Basically, just start MC, start playing a list and then about 30 minutes later, the above.  No JRWorkers running.

[bob]

Huge mem leak of some kind:

Code: [Select]

The process mediacenter24 (with pid 1966) is using approximately 2.4 GB of memory.
It is using 2.4 GB privately, and a further 12.7 MB that is, or could be, shared with other programs.
Dividing up the shared memory between all the processes sharing that memory we get a reduced shared memory usage of 2.9 MB. Adding that to the private usage, we get the above mentioned total memory footprint of 2.4 GB.
Did not see this in 26.  Basically, just start MC, start playing a list and then about 30 minutes later, the above.  No JRWorkers running.

What kind of material are you playing?
What's the output device?
Also, what are your audio settings for track transitions? (gapped, gapless, and the rest).
What command are you using to provide the message about memory usage?

[Mike Noe]

1.  Playing APEs and FLACs
2.  Alsa (Pulse Audio Sound Server)
3.  X-fade aggressive (4s), Do not play silence/Use gapless for seq album tracks
4.  Using Plasma5's KSysguard system monitor (I did not try top)

I notice also a not-insignficant I/O rate as well, GBs of data over a few hour period...weird

[mwillems]

1.  Playing APEs and FLACs
2.  Alsa (Pulse Audio Sound Server)
3.  X-fade aggressive (4s), Do not play silence/Use gapless for seq album tracks
4.  Using Plasma5's KSysguard system monitor (I did not try top)

I notice also a not-insignficant I/O rate as well, GBs of data over a few hour period...weird

Is this local playback or is there a client-server setup involved; I've still got a pretty big memory leak on linux clients when "auto-sync with server" is enabled.

[Mike Noe]

Technically local playback, but I do have 5 zones, so technically it's not just the Alsa device.  1 zone uses a USB-DAC, another zone uses the HDMI device, another zone uses the mobo HD audio device. 

In addition, I have a separate process that runs for each zone on my LAN (there's also an MC server on a laptop) that exposes an MPRIS interface for that zone, which I then can access via my android device with KDE Connect.

Then, there is a plasmoid running that polls each zone for playback status (QML/JS type thing) every few seconds.  So there is more than a little activity against the MCWS.

BUt, I've had this going for quite some time (way back to early 23.x) and this huge memory alloc just started happening with 24.28. 

[bob]

Technically local playback, but I do have 5 zones, so technically it's not just the Alsa device.  1 zone uses a USB-DAC, another zone uses the HDMI device, another zone uses the mobo HD audio device. 

In addition, I have a separate process that runs for each zone on my LAN (there's also an MC server on a laptop) that exposes an MPRIS interface for that zone, which I then can access via my android device with KDE Connect.

Then, there is a plasmoid running that polls each zone for playback status (QML/JS type thing) every few seconds.  So there is more than a little activity against the MCWS.

BUt, I've had this going for quite some time (way back to early 23.x) and this huge memory alloc just started happening with 24.28.

I ran one all night and the memory usage didn't budge.

Can you try changing the Audio settings to gapless track transitions and uncheck the don't play silence box.

[bob]

Is this local playback or is there a client-server setup involved; I've still got a pretty big memory leak on linux clients when "auto-sync with server" is enabled.

Just linux clients?

[mwillems]

Just linux clients?

I haven't seen it on any windows clients, and the server is on windows and has perfectly stable memory use.

[bob]

I haven't seen it on any windows clients, and the server is on windows and has perfectly stable memory use.

Ok, thanks, does that seem to be a new thing as well?

[mwillems]

Ok, thanks, does that seem to be a new thing as well?

No its at least a few months old, maybe older; its the same issue I described in this thread: https://yabb.jriver.com/interact/index.php/topic,115120.0.html

[bob]

No its at least a few months old, maybe older; its the same issue I described in this thread: https://yabb.jriver.com/interact/index.php/topic,115120.0.html

Thanks for reminding me.
Do you have memory playback enabled?

[Mike Noe]

I ran one all night and the memory usage didn't budge.

Can you try changing the Audio settings to gapless track transitions and uncheck the don't play silence box.

Think you're on to something.  Changed the transitions to be as you've requested, stopped and started MC, and now, after about an hour, I'm still at "normal" mem usages (~20 MCWS connections, playback on 2 zones)

Do not play silence:  Unchecked
Use gapless for seq... :  Checked
Use gapless for manual... :  Checked

[mwillems]

Thanks for reminding me.
Do you have memory playback enabled?

No, memory playback is disabled. 

[bob]

Think you're on to something.  Changed the transitions to be as you've requested, stopped and started MC, and now, after about an hour, I'm still at "normal" mem usages (~20 MCWS connections, playback on 2 zones)

Do not play silence:  Unchecked
Use gapless for seq... :  Checked
Use gapless for manual... :  Checked

What was the track transition set to originally? Gapped or crossfade?

[bob]

No, memory playback is disabled.

I'm not seeing excessive sockets opened, can you post an example?

[mwillems]

I'm not seeing excessive sockets opened, can you post an example?

Sorry, the sockets were a red-herring I think (I mentioned that down-thread here with some lsof output and other testing results: https://yabb.jriver.com/interact/index.php/topic,115120.msg796478.html#msg796478); one of my pis had lots of extra sockets, but my desktop linux client (which runs amd64) doesn't see extra sockets, but both do see gradually increasing memory usage.  So I think the sockets are/were an unrelated issue with my Pi, but the memory leak is real.

So the issue as it currently stands is that all of my Linux clients gradually increase the amount of memory used while "auto-sync with server" is enabled, and don't do so when its disabled (or when there's no client/server connection).  The memory growth increases in speed when playing files (predictably on track change), but increases gradually even at idle.  The memory growth continues until available memory is exhausted at which point MC crashes.  When disabling auto-sync with server, memory growth stops but existing extra memory is not relinquished.  This is reproducible with one client and one server, or with many clients.

I'm not sure how to post an example that would demonstrate this, but let me know what else might be useful to know.

[bob]

Sorry, the sockets were a red-herring I think (I mentioned that down-thread here with some lsof output and other testing results: https://yabb.jriver.com/interact/index.php/topic,115120.msg796478.html#msg796478); one of my pis had lots of extra sockets, but my desktop linux client (which runs amd64) doesn't see extra sockets, but both do see gradually increasing memory usage.  So I think the sockets are/were an unrelated issue with my Pi, but the memory leak is real.

So the issue as it currently stands is that all of my Linux clients gradually increase the amount of memory used while "auto-sync with server" is enabled, and don't do so when its disabled (or when there's no client/server connection).  The memory growth increases in speed when playing files (predictably on track change), but increases gradually even at idle.  The memory growth continues until available memory is exhausted at which point MC crashes.  When disabling auto-sync with server, memory growth stops but existing extra memory is not relinquished.  This is reproducible with one client and one server, or with many clients.

I'm not sure how to post an example that would demonstrate this, but let me know what else might be useful to know.

How are your playback track transitions set?
I'm wondering if that's the same issue as the one above?

[mwillems]

How are your playback track transitions set?
I'm wondering if that's the same issue as the one above?

Currently I have:

Do not play silence:  Checked
Use gapless for seq... :  Checked
Use gapless for manual... :  Unchecked

I'll try changing those settings to the recommended and see what happens.  I'm not sure how "do not play silence" got checked in any case, I always try to uncheck it on my instances.

[bob]

Currently I have:

Do not play silence:  Checked
Use gapless for seq... :  Checked
Use gapless for manual... :  Unchecked

I'll try changing those settings to the recommended and see what happens.  I'm not sure how "do not play silence" got checked in any case, I always try to uncheck it on my instances.

And what's the actual track transition type (gapped, gapless ,etc)?

[mwillems]

And what's the actual track transition type (gapped, gapless ,etc)?

Cross-fade aggressive (which I think is, or was once, the default?)

[bob]

Cross-fade aggressive (which I think is, or was once, the default?)

I'm guessing that's the issue.
Setting up a test for the weekend...

[Mike Noe]

Cross-fade aggressive (which I think is, or was once, the default?)

Same here...

[Rizlaw]

When I installed MC 24.0.26, the update noted that restoring the license would be required. OK, I did that.

Then I received the 24.0.28 update notice from the MC "latest" repository. I installed it without issue and used it for several days with no license notifications.

Today, after an Ubuntu kernel update from 4.15.20 to 4.15.22, I now get an MC 24.0.28 "restore your license" message.

I thought this was fixed with MC 24.0.26?  ?

[Awesome Donkey]

Can confirm with the kernel update, MC lost its registration.

[rcblackwell]

...... I received the 24.0.28 update notice from the MC "latest" repository. I installed it without issue.....

....I now get an MC 24.0.28 "restore your license" message.

I received the same message after upgrading as noted by Rizlaw

Robert

[geier22]

I think it must be something else. I have a triple boot system (KDE / XFCE / LXQT - all Debian Buster). There has been no kernel update in buster in the last few days (4.16.0-1-amd64 # 1 SMP Debian 4.16.5-1 (2018-04-29) x86_64 GNU / Linux). In all three systems I have lost the license for MC 24.0.28 in the last 3 days after an update and reboot.

[Awesome Donkey]

Yep, must be a MC bug because registration was lost in my Arch Linux, Debian and Linux Mint VMs without performing any system update.

[Awesome Donkey]

Finally had the registration to be lost in my main Arch Linux install. Is it related to kernel updates? As far as I know, there might be mitigations to Spectre variant 4 in those kernel updates.

Nonetheless, all my Linux installs have lost the registration now. So I've reactivated all of them, and we'll see if it fails again.

[bob]

If you hadn't re-entered the license in 28.0.26 intel, you should have to do it in 24.0.28.
Otherwise it should be OK.
If someone can take a log on the machine that's forced you to change AFTER also doing a change in 24.0.26 and send it to me I'd appreciate it.
It's possible a kernel update could do this, I'll know more when I see a log.

bob (at) jriver (dot) com

[blgentry]

I've installed 24.0.28 and (separately and removed it) 24.0.45.  Both exhibit the following behavior:

After launching correctly, I cancel the "go find it all" import and select File > Library > Restore Library .  Then navigate to a backup made with MC23 for Linux minutes ago.  Almost immediately, at the terminal I launched MC from, it get:

Code: [Select]

Gtk-Message: GtkDialog mapped without a transient parent. This is discouraged.

[1]+  Segmentation fault      mediacenter24

...and MC disappears.

Any ideas?

Brian.

[bob]

I've installed 24.0.28 and (separately and removed it) 24.0.45.  Both exhibit the following behavior:

After launching correctly, I cancel the "go find it all" import and select File > Library > Restore Library .  Then navigate to a backup made with MC23 for Linux minutes ago.  Almost immediately, at the terminal I launched MC from, it get:

Code: [Select]

Gtk-Message: GtkDialog mapped without a transient parent. This is discouraged.

[1]+  Segmentation fault      mediacenter24

...and MC disappears.

Any ideas?

Brian.

The gtk message doesn't appear to be important, I made a build to test this with another person that was having the same problem and it didn't help.
I was not able to duplicate the issue here, it appears to be a bug in cinnamon. Are you using that?

[blgentry]

I'm quite impressed with your remote troubleshooting!

Yes, I've been using cinnamon for quite a long time on my Linux desktop here at work.  I just switched to XFCE as an experiment.  Now MC seems to work as expected.

Under cinnamon MC did other weird stuff, like the menus all appeared in an animated way, seeming to start small, and quickly growing.  This was pretty weird looking when traversing down several levels of menus.  That behavior is gone now with XFCE as well.

I'm not sure if I can live with XFCE full time just yet, but I'm going to give it a try.

Thanks!

Brian.

[bob]

I'm quite impressed with your remote troubleshooting!

Yes, I've been using cinnamon for quite a long time on my Linux desktop here at work.  I just switched to XFCE as an experiment.  Now MC seems to work as expected.

Under cinnamon MC did other weird stuff, like the menus all appeared in an animated way, seeming to start small, and quickly growing.  This was pretty weird looking when traversing down several levels of menus.  That behavior is gone now with XFCE as well.

I'm not sure if I can live with XFCE full time just yet, but I'm going to give it a try.

Thanks!

Brian.

I wonder if it's an issue with the compositor?
If the video card is OpenGL capable and Dri is working properly I think the compositor will be enabled automatically.
If you have an NVidia card, the only time I've gotten this work work properly is with the Nvidia supplied driver (not Nouveau or the open source driver).
You might try cinnamon with the compositor turned off (if that's an option somewhere).

[Awesome Donkey]

Hmm, wonder if I can get it to happen in my Linux Mint 19 VM (its DE is Cinnamon). Somebody should check GNOME too, as Cinnamon is a fork of GNOME 3 and the issue might happen there too. Unfortunately, I can't reproduce it in my VMs thus far. The *only* issue(s) with GNOME I've noticed are a) the save file dialog appears behind the main MC window (e.g. when doing a library backup) and b) if using Dash To Dock (or using Ubuntu 18.04 with its default dock) if you start MC and minimize, for about 30 seconds MC appears busy and if you click on MC's icon in the dock it'll launch a second instance. If you wait 30 seconds and try this, it'll restore the MC window instead of launching a new instance. Something I've noticed for a couple years now.

If you're using a Nvidia 9xx or 10xx series GPU, you have to use the proprietary Nvidia driver, as the open-source Nouveau/Mesa support for these GPUs isn't very good (thanks to Nvidia either not helping or giving the firmware blobs needed).

[blgentry]

I'm using the video card built in to the motherboard of an inexpensive Dell desktop.  Looks like it's an Intel graphics chip.  Xeon E3-1200 integrated, according to lspci.

I might mess around with the cinnamon settings in a bit. 

Brian.

[Awesome Donkey]

What distro are you using? Also can you find out which version of Cinnamon its running (via system info or something like that within the distro)?

[Awesome Donkey]

Okay, got it to segfault when restoring a library backup in Linux Mint 19. It didn't do it the first two times I did it, but if I switch between two backups restoring one after the other, it eventually just gave out.

Code: [Select]

Gtk-Message: 10:48:53.691: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:49:01.508: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:49:47.518: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:50:01.496: GtkDialog mapped without a transient parent. This is discouraged.
Segmentation fault (core dumped)
I'll see if I can do this in Ubuntu with GNOME next.

EDIT: Tried switching between a library backup from Ubuntu and a library backup from Linux Mint in Ubuntu, did the restore library switch 10 times and no crash. If I repeat this in Linux Mint, MC will crash with a segfault after restoring one of the libraries a few times, so it seems GNOME isn't affected and Cinnamon is. I'll try KDE Plasma next.

[Hendrik]

I'm installing Linux Mint with Cinnamon in a VM just now, lets see if a VM is enough to reproduce it.

[bob]

Okay, got it to segfault when restoring a library backup in Linux Mint 19. It didn't do it the first two times I did it, but if I switch between two backups restoring one after the other, it eventually just gave out.

Code: [Select]

Gtk-Message: 10:48:53.691: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:49:01.508: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:49:47.518: GtkDialog mapped without a transient parent. This is discouraged.
Gtk-Message: 10:50:01.496: GtkDialog mapped without a transient parent. This is discouraged.
Segmentation fault (core dumped)
I'll see if I can do this in Ubuntu with GNOME next.

EDIT: Tried switching between a library backup from Ubuntu and a library backup from Linux Mint in Ubuntu, did the switch 10 times and no crash. If I repeat this in Linux Mint, MC will crash with a segfault, so it seems GNOME isn't affected.

I can give you a link to try that will eliminate the GtkDIalog message but I don't think that's it since the issue still appeared after that with the other person I was testing with.

We did find something interesting a couple of days ago with orphaned entries in the /run/shm directory.
With MC NOT running you might want to try removing all of the
JRCrashInfo
MJCPI
JRiver Process v24
Media Center 24

entries and try again.

[bob]

I'm installing Linux Mint with Cinnamon in a VM just now, lets see if a VM is enough to reproduce it.

I tried this on Ubuntu 18.04 with Cinnamon and it worked fine.

[Awesome Donkey]

If I recall correctly, Ubuntu ships with an older version of Cinnamon whereas Linux Mint ships the latest 3.8.8 version of Cinnamon. Also all distros give off the Gtk-Message warning (even those that don't crash) so I'll just ignore that for now.

I can reproduce this with Cinnamon 3.8.8 in Linux Mint 19. Only takes a couple library restores to do it. Also I just tried this in KDE Plasma, and surprise surprise, no crash after restoring library backups 10 times.

I'm installing Linux Mint with Cinnamon in a VM just now, lets see if a VM is enough to reproduce it.

Should be, I can reproduce it in my Linux Mint 19 VMware VM. Just make a couple library backups and keep restoring one after the other until it goes belly up.

Perhaps we need a topic to report DE-related issues with MC. I know of a couple with GNOME and KDE Plasma, but oddly when I used Cinnamon some months back it worked perfectly fine there.

[bob]

If you want to try it with the build that removes the gtk warning:

http://files.jriver.com/mediacenter/test/MediaCenter-24.0.45-2-amd64.deb

Note that this has the new browser window and will do a bug chromium download. (and will likely need a license restore after running build 28).

[Hendrik]

Definitely can reproduce it, but not with restore library, but as with the previous report using Import Single Folder.

[bob]

Definitely can reproduce it, but not with restore library, but as with the previous report using Import Single Folder.

Did you try the build above?

[blgentry]

What distro are you using? Also can you find out which version of Cinnamon its running (via system info or something like that within the distro)?

I'm running Debian with Cinnamon version 2.2.16-5.  I've been running MC22 and MC23 on this computer, in this configuration for something like a year.  I just decided to upgrade it to MC24 today.

Brian.

[blgentry]

We did find something interesting a couple of days ago with orphaned entries in the /run/shm directory.
With MC NOT running you might want to try removing all of the
JRCrashInfo
MJCPI
JRiver Process v24
Media Center 24

entries and try again.

I just switched back to cinnamon, and tried again.  I removed the above entries.  Next I started by clearing my MC24 library, and then did a restore.  Works totally normally!

Now this isn't identical to what I was doing before, because I was in "MC initial state with no library and wanting one", where it does the countdown thing.  Now of course, I can't get back to that state.

Anyway, I'm back to my old window manager, and MC24 works.

Thanks everyone! 

Brian.