LinuxMusicians

With kernel 4.4.0 on my Xubuntu 16.10 everything is fine:

[…]
connect(23, {sa_family=AF_LOCAL, sun_path="/run/user/1000/pulse/native"}, 110) = 0
write(22, "W", 1)                       = 1
clock_gettime(CLOCK_MONOTONIC, {52, 963145311}) = 0
write(22, "W", 1)                       = 1
futex(0x7ff40515d7b0, FUTEX_UNLOCK_PI_PRIVATE, 33) = 0
futex(0x7ff40515d7e4, FUTEX_WAIT_REQUEUE_PI_PRIVATE, 9, NULL, 0x7ff40515d7b0) = 0
write(22, "W", 1)                       = 1
clock_gettime(CLOCK_MONOTONIC, {52, 963516310}) = 0
write(22, "W", 1)                       = 1
futex(0x7ff40515d7b0, FUTEX_UNLOCK_PI_PRIVATE, 33) = 0
futex(0x7ff40515d7e4, FUTEX_WAIT_REQUEUE_PI_PRIVATE, 11, NULL, 0x7ff40515d7b0) = 0
write(22, "W", 1)                       = 1
clock_gettime(CLOCK_MONOTONIC, {52, 963686445}) = 0
write(22, "W", 1)                       = 1
futex(0x7ff40515d7b0, FUTEX_UNLOCK_PI_PRIVATE, 33) = 0
futex(0x7ff40515d7e4, FUTEX_WAIT_REQUEUE_PI_PRIVATE, 13, NULL, 0x7ff40515d7b0) = 0
write(22, "W", 1)                       = 1
close(25)                               = 0
close(24)                               = 0
munmap(0x7ff3f65fe000, 67112960)        = 0
write(22, "W", 1)                       = 1
close(23)                               = 0
munmap(0x7ff3fa5ff000, 67112960)        = 0
unlink("/dev/shm/pulse-shm-4254065881") = 0
futex(0x7ff40515d7b0, FUTEX_UNLOCK_PI_PRIVATE, 33) = 0
write(17, "[2023.10.15 12:16:44] Audio Capt"..., 126) = 126 
[…]

But with kernel 5.15.134 Telegram hangs before it can display its UI:

[…]
connect(26, {sa_family=AF_LOCAL, sun_path="/run/user/1000/pulse/native"}, 110) = 0
sendto(25, "W", 1, MSG_NOSIGNAL, NULL, 0) = -1 ENOTSOCK (Socket operation on non-socket)
write(25, "W", 1)                       = 1
clock_gettime(CLOCK_MONOTONIC, {50, 589784800}) = 0
write(25, "W", 1)                       = 1
futex(0x7f4ba8d4ecf0, FUTEX_UNLOCK_PI_PRIVATE, 33) = -1 EPERM (Operation not permitted)
futex(0x7f4ba8d4ed24, FUTEX_WAIT_REQUEUE_PI_PRIVATE, 1, NULL, 0x7f4ba8d4ecf0) = ? ERESTARTNOINTR (To be restarted)
=> at this point Telegram hangs

Any idea? Is there a new kernel variable with late v5 kernels that has to be set for backwards compatibility? New permissions for my user that have to be set?

I don't use *buntu but the telegram-desktop version from Debian repo, currently v.4.6.5, is working fine here in Debian Bullseye with Liquorix kernel 6.4.x
Are you using a recent version of Telegram?

sunrat wrote: ↑Sun Oct 15, 2023 10:51 am

I don't use *buntu but the telegram-desktop version from Debian repo, currently v.4.6.5, is working fine here in Debian Bullseye with Liquorix kernel 6.4.x
Are you using a recent version of Telegram?

Yes. 4.9.9 from telegram.org. It has its own updater.

But the problem occurs with 4.2.4 and 4.5.3 as well.

And this is on the same installation where it works with the older kernel but not with the newer one?

autostatic wrote: ↑Sun Oct 15, 2023 1:19 pm

And this is on the same installation where it works with the older kernel but not with the newer one?

Yep. When I boot the v4 kernel everything works as expected, when I boot the v5 kernel I get the problem with Telegram and Pulseaudio.

That's weird! Any difference in file permissions of /run/user/1000/pulse/native between those two kernels?

autostatic wrote: ↑Sun Oct 15, 2023 3:45 pm

That's weird! Any difference in file permissions of /run/user/1000/pulse/native between those two kernels?

No difference.

Kernel v4:

drwx------ 2 hm hm 80 Okt 15 18:06 /run/user/1000/pulse
srwxrwxrwx 1 hm hm 0 Okt 15 18:06 /run/user/1000/pulse/native

Kernel v5:

drwx------ 2 hm hm 80 Okt 15 18:10 /run/user/1000/pulse
srwxrwxrwx 1 hm hm 0 Okt 15 18:10 /run/user/1000/pulse/native

The plot thickens. You might think this is a permissions error when looking at the strace output. And also, PA is a user space app, a different kernel shouldn't make a difference. Unless the newer kernel has difficulties with your audio devices. aplay -l looks the same too with both kernels?

autostatic wrote: ↑Sun Oct 15, 2023 4:22 pm

The plot thickens. You might think this is a permissions error when looking at the strace output. And also, PA is a user space app, a different kernel shouldn't make a difference. Unless the newer kernel has difficulties with your audio devices. aplay -l looks the same too with both kernels?

Pulseaudio outputs to the already running jack2 and isn't aware which hardware is attached.

I guess I have to compile a recent Pulseaudio to find out more.

EDIT:
No difference with a more recent version pulseaudio-14.2. So I give up for now.

bluebell wrote: ↑Sun Oct 15, 2023 10:30 am

[...] kernel 4.4.0 on my Xubuntu 16.10

That's a very, very old distro. Why would you want to update the kernel if you do not have problems?

Linuxmusician01 wrote: ↑Mon Oct 16, 2023 9:26 am

bluebell wrote: ↑Sun Oct 15, 2023 10:30 am

[...] kernel 4.4.0 on my Xubuntu 16.10

That's a very, very old distro. Why would you want to update the kernel if you do not have problems?

For @Drumfix' ua101 replacement driver for my MOTU UltraLite AVB (doesn't compile with a v4 kernel) and improvements for my Scarlett 18i20.

Works fine.

Of course there are some problems. I can compile Qtractor either with CLAP or gtkmm support. gcc5 compiles programs using gtkmm, gcc7 and 9 compile programs using CLAP. I can't have both. Earlier or later I'll have to install a new system and the game will begin again.

bluebell wrote: ↑Sun Oct 15, 2023 10:30 am

With kernel 4.4.0 on my Xubuntu 16.10 everything is fine:

Running anything network-related on so old system feels pretty risky for me. No updates for last 6 years

tavasti wrote: ↑Mon Oct 16, 2023 11:01 am

bluebell wrote: ↑Sun Oct 15, 2023 10:30 am

With kernel 4.4.0 on my Xubuntu 16.10 everything is fine:

Running anything network-related on so old system feels pretty risky for me. No updates for last 6 years

It must be totally secure because I haven't got security updates for years.

Even my private servers connected with a packet filter are not up to date because I update only to fix relevant security holes that can't be fixed by stopping or deinstalling services that are not needed. Result: My machines weren't vulnerable by the glibc bug because the glibc wasn't recent enough to have this bug.

I remember old times when I was a systems programmer for IBM mainframes. IBM field technicians told us "Don't fix what ain't broken."

Times have changed. If I still had to run servers for a living I'd join the "recent updates are more important than running systems" religion for my employer.

Totally unrelated but a free Ubuntu One account allows you to use ESM on 5 machines. This will keep your 16.04 machine secure until 2026.

Edit: you're running a non-LTS version, 16.10. No ESM for those versions

autostatic wrote: ↑Mon Oct 16, 2023 12:04 pm

Totally unrelated but a free Ubuntu One account allows you to use ESM on 5 machines. This will keep your 16.04 machine secure until 2026.

Edit: you're running a non-LTS version, 16.10. No ESM for those versions

Sorry, it's 16.04 LTS. I think about it for my mailserver. Updating from 12.04 via 14.04 to 16.04 was a pain in the ass. Don't feel like updating to 18.x or 20.x in the next months.