Hi all.
After upgrading to fc22 during summer, I'm having issues with 32bit jack
clients running in my 64bit jackd.
I'd be thankful for any clues to what's going on, and ways to debug
this. And if others would care to check, and confirm/disprove the same
behavior, please tell me what you get.
Here's what happens:
The 32bit clients connect well, but the callbacks never seem to return,
making jackd send out a steady stream of xruns and errors:
JackEngine::XRun: client = simple_client was not finished, state = Triggered
I'm checking with jack2 installed (jack-1.9.10), with .i686 and .x86_64
versions installed using the standard fedora packages.
The same 32bit clients work well in fc21, fc20, and ubuntu 14.04 (in a
vm on the same hardware). Afaik, nothing has changed vs. the -mixed
flag in jackd, and would expect 32bit clients to run well inside a 64bit
jackd.
Checking with example-clients/simple-client.c from the jack sources:
$ gcc -m32 -g -O -lm -ljack simple_client.c -o simple_client
$ ./simple_client
- results in the noted behavior. Without -m32 everything's fine.
Thanks,
-anders
On a Linux 4.2 system with realtime scheduling controlled by cgroups, what is
the current recommended strategy to get jackd working as a non-root user?
There was an old wiki page,
https://github.com/jackaudio/jackaudio.github.com/wiki/Cgroups
describing some built-in support (e.g. --with-default-cgroup configuration
switch). But none of this is in the current git HEAD. (I am using jack2
1.9.11.)
And a somewhat related question: Is there currently support for running
jack as root and allowing client programs with multiple UIDs to connect? I
tried following an old recipe using JACK_PROMISCUOUS_SERVER, but it didn't
seem to make any difference; running jackd as root, only root can connect.
Regards,
Mark
On Sun, Aug 23, 2015 at 03:49:26PM +0000, Fons Adriaensen wrote:
> I also completely replaced the code in Jack1 that
> calculates the proper running order of clients.
> The previous algorithm failed to do this in some
> cases. It could not be 'fixed' easily as it was
> basically using the wrong algorithm.
>
> Affected files are
>
> modified: include/engine.h
> modified: include/internal.h
> modified: jackd/clientengine.c
> modified: jackd/clientengine.h
> modified: jackd/engine.c
>
> There seems to be no interest from the Jack devs,
> but if anybody wants to test this I can either
> provide the modified files or a patch against
> git commit 5af5815c47630b77cc71c91a460f8aa398017cf7
> (current HEAD).
Not that I consider myself a jack-dev anymore, but how about you just
share out your patch on jack-devel or as a pull-request on github?
My impression is that somebody is just going to merge it. ;)
Cheers
--
mail: adi(a)thur.de http://adi.thur.de PGP/GPG: key via keyserver
Hi!
This just caught my attention. I didn't look into it, yet, but maybe
somebody has some cycles to investigate.
Cheers
-------- Forwarded Message --------
Subject: Bug#786737: jackd1: crashes with -n option specified
Resent-Date: Mon, 25 May 2015 02:57:01 +0000
Resent-From: Frank Heckenbach <f.heckenbach(a)fh-soft.de>
Resent-To: debian-bugs-dist(a)lists.debian.org
Resent-CC: Debian Multimedia Maintainers
<pkg-multimedia-maintainers(a)lists.alioth.debian.org>
Date: Mon, 25 May 2015 04:53:18 +0200
From: Frank Heckenbach <f.heckenbach(a)fh-soft.de>
Reply-To: Frank Heckenbach <f.heckenbach(a)fh-soft.de>, 786737(a)bugs.debian.org
To: submit(a)bugs.debian.org
Package: jackd1
Version: 1:0.124.1+20140122git5013bed0-3
Severity: normal
Tags: upstream patch
When the "-n" option is given, jackd crashes when accessing
properties (which it seems to do implicitly for any client, e.g.
jack_lsp).
To reproduce:
jackd -nfoo -dalsa
or
JACK_DEFAULT_SERVER=bar jackd -nfoo -dalsa
While jackd is running, repeat this:
JACK_DEFAULT_SERVER=foo jack_lsp
After a few tries, jackd crashes:
/dev/shm/jack-0/default/__db.001: No such file or directory
cannot open DB environment: No such file or directory
Segmentation fault
I found several problems in the code:
- jack_property_init():
Returns early if db_env != NULL. However, on failure later returns
with db_env != NULL, but db == NULL. Callers assume db != NULL
after it returns.
Patch below. This fixes at least the segfault, but the error
message remains (and whatever consequences it may have, I don't
know).
- 7 places in metadata.c:
if (jack_property_init (NULL)) {
If NULL is passed for server_name, of course it won't use the
actual server name.
How is this supposed to work at all? Was it ever tested?
A fix should consist of passing the actual server name from the
callers.
--- libjack/metadata.c
+++ libjack/metadata.c
@@ -45,11 +45,11 @@
/* idempotent */
- if (db_env) {
+ if (db) {
return 0;
}
- if ((ret = db_env_create(&db_env, 0)) != 0) {
+ if (!db_env && (ret = db_env_create(&db_env, 0)) != 0) {
jack_error ("cannot initialize DB environment: %s\n",
db_strerror(ret));
return -1;
}
_______________________________________________
pkg-multimedia-maintainers mailing list
pkg-multimedia-maintainers(a)lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-multimedia-main…
On 08/19/2015 10:15 AM, Matt Flax wrote:
> Good news,
>
> It turns out these problems were because shm was on a very space
> restricted device ... I moved shm to the tmp directory and it is now
> functioning as it should.
>
> jack1 on MIPS arch. is tested working.
> I ran jackd 1 using 2ms latency, no xruns. That is with a 48kHz sample
> rate.
>
> I played from my desktop using netjack one and I have seen the I2S data
> flowing out ... nice first test. When I get an audio codec connected I
> will do a listening test.
>
> Matt
Awesome! What kind of board are you using by the way?
Jeremy
Hi Jack people,
I am doing a lot of testing of various types of Windows audio software
at the moment. ReWire is good but Jack would make a nice alternative. I
tested in Windows Vista and got Jack Audio running, but it crashed a few
times and I got very crackly audio. After moving to my Windows 7 64bit
dual boot installation (I eventually discovered the custom/hacked PAX
drivers for my X-Fi card to replace buggy Creative drivers) I found Jack
more stable but still get this crackling in the audio - this is not
present in any other audio from my system, MME, ASIO, games, ReWire,
etc.
So how do I get involved with the Windows version ? I can put in bug
reports and do testing. I have been known to fire up windows compilers
(after eating the right kind of mushrooms, ya know how it is) but I'm
not an experienced coder. Frankly I wonder what is causing the crackling
because the rest of Jack in Windows seems to be working and stable.
DJ Barney
--
~~~
http://djbarney.org
Hi there,
I am wanting to run jackd on low resourced/power embedded systems
(running Linux).
When I run jackd on these systems it tries to lock down way to much
memory ... much more then is available.
Is the amount of memory locked down by jackd configurable ? Or is it an
operating system set thing ?
Does anyone have any constructive ideas or suggestions for running jackd
on these low resource embedded systems ?
thanks
Matt
Hi all,
as in through the hottest of summers--as southerners can't even
wait--here comes part one:
QjackCtl 0.4.0 (summer'15) is out!
though aside that everybody knows this already,
QjackCtl is a (maybe not so any more but) simple Qt [3] application
to control the JACK [2] sound server, for the Linux Audio infrastructure.
website:
http://qjackctl.sourceforge.net
downloads:
http://sourceforge.net/projects/qjackctl/files
- source tarball:
http://download.sourceforge.net/qjackctl/qjackctl-0.4.0.tar.gz
- source package:
http://download.sourceforge.net/qjackctl/qjackctl-0.4.0-23.rncbc.suse132.sr…
- binary packages:
http://download.sourceforge.net/qjackctl/qjackctl-0.4.0-23.rncbc.suse132.i5…http://download.sourceforge.net/qjackctl/qjackctl-0.4.0-23.rncbc.suse132.x8…
Change-log:
- Some windows fixes added (patch by Kjetil Matheussen, thanks).
- Most advanced Setup/Settings are moved into new Setup/Advanced
settings tab; limit range for the real-time priority setting, now having
6 as absolute minimum valid value (after patches by Robin Gareus, thanks).
- A new top-level widget window geometry state save and restore
sub-routine is now in effect (EXPERIMENTAL)
- Delayed geometry setup for widget windows upon startup has been
deprecated and scrapped altogether.
- Setup/settings dialog tab is going into some layout changes; also got
rid of old patchbay auto-refresh timer cruft, which was previously
hidden/disabled.
- New socket names are now automatically inferred from selected client
names while on the Patchbay widget, Socket dialog.
- Fixed for some strict tests for Qt4 vs. Qt5 configure builds.
- German (de) translation update (by Guido Scholz, thanks).
License:
QjackCtl stands free, still open-source software, distributed under
the terms of the GNU General Public License (GPL [4]) version 2 or later.
Weblog (upstream support):
http://www.rncbc.org
See also:
http://www.rncbc.org/drupal/node/912
References:
[1] QjackCtl - A JACK Audio Connection Kit Qt GUI Interface
http://qjackctl.sourceforge.net
[2] JACK Audio Connection Kit
http://jackaudio.org
[3] Qt framework, C++ class library and tools for
cross-platform application and UI development
http://qt.io/
[4] GPL - GNU General Public License
http://www.gnu.org/copyleft/gpl.html
Enjoy && have a whole hot'ta Summer'15 fun!
--
rncbc aka. Rui Nuno Capela
--
rncbc aka. Rui Nuno Capela