[Jack-Devel] message 7

Prev	Next Index
Date	Mon, 25 Apr 2011 14:10:26 +0100
From	david lees <[hidden] at ntlworld dot com>
To	[hidden] at lists dot jackaudio dot org
On 04/24/2011 11:06 PM, [hidden] wrote:
> Send Jack-Devel mailing list submissions to
> 	[hidden]
>
> To subscribe or unsubscribe via the World Wide Web, visit
> 	http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
> or, via email, send a message with subject or body 'help' to
> 	[hidden]
>
> You can reach the person managing the list at
> 	[hidden]
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Jack-Devel digest..."
>
>
> Today's Topics:
>
>     1. gdb-ing jack apps? (Dan Muresan)
>     2. Re: gdb-ing jack apps? (James Morris)
>     3. Re: gdb-ing jack apps? (Dan Muresan)
>     4. Re: gdb-ing jack apps? (Paul Davis)
>     5. Re: gdb-ing jack apps? (Dan Muresan)
>     6. RFC: fix for missing port registration notifications	when
>        switching master (Nedko Arnaudov)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sun, 24 Apr 2011 10:25:11 +0300
> From: Dan Muresan<[hidden]>
> To: [hidden]
> Subject: [Jack-Devel] gdb-ing jack apps?
> Message-ID:<[hidden]>
> Content-Type: text/plain; charset=ISO-8859-1
>
> Hi,
>
> I've been unable to debug my jack app in gdb. Take the auto-server
> case. Whenever gdb stops the program due to a signal (either ^C or a
> segfault) an endless stream of
>
> subgraph starting at<client>  timed out (subgraph_wait_fd=14, status =
> 0, state = Triggered, pollret = 0 revents = 0x0)
>
> gets printed, and I have to kill gdb from another shell.
>
> If I start the server from qjackctl, the same process crashes qjackctl.
>
> I've tried exiting the process thread on lost frames, but it doesn't help.
>
> I've seen reports of jackd and gdb on the net, so there has to be a
> way. What could be the problem? I am of course not expecting to stop a
> jack app and then continue, but at least I want to be able to see
> backtraces, inspect variables etc in gdb.
>
> Thanks,
> Dan
>
>
> ------------------------------
>
> Message: 2
> Date: Sun, 24 Apr 2011 12:05:28 +0100
> From: James Morris<[hidden]>
> To: Dan Muresan<[hidden]>
> Cc: [hidden]
> Subject: Re: [Jack-Devel] gdb-ing jack apps?
> Message-ID:<BANLkTiku3q=HehFGT4w9r3gTjZ+HL=[hidden]>
> Content-Type: text/plain; charset=ISO-8859-1
>
> On 24 April 2011 08:25, Dan Muresan<[hidden]>  wrote:
>> Hi,
>>
>> I've been unable to debug my jack app in gdb. Take the auto-server
>> case. Whenever gdb stops the program due to a signal (either ^C or a
>> segfault) an endless stream of
>>
>> subgraph starting at<client>  timed out (subgraph_wait_fd=14, status =
>> 0, state = Triggered, pollret = 0 revents = 0x0)
>>
>> gets printed, and I have to kill gdb from another shell.
>>
>> If I start the server from qjackctl, the same process crashes qjackctl.
>>
>> I've tried exiting the process thread on lost frames, but it doesn't help.
>>
>> I've seen reports of jackd and gdb on the net, so there has to be a
>> way. What could be the problem? I am of course not expecting to stop a
>> jack app and then continue, but at least I want to be able to see
>> backtraces, inspect variables etc in gdb.
> One way is to get core dumps from the app (ulimit -c unlimited) and
> then gdb ./myapp core, and then in gdb: thread apply all bt.
>
> You can also use the dummy driver (instead of, for example Alsa) with
> a long --timeout value to be able to step through the code.
>
> James.
>
>
> ------------------------------
>
> Message: 3
> Date: Sun, 24 Apr 2011 14:10:40 +0300
> From: Dan Muresan<[hidden]>
> To: James Morris<[hidden]>
> Cc: [hidden]
> Subject: Re: [Jack-Devel] gdb-ing jack apps?
> Message-ID:<[hidden]>
> Content-Type: text/plain; charset=ISO-8859-1
>
>> One way is to get core dumps from the app (ulimit -c unlimited) and
>> then gdb ./myapp core, and then in gdb: thread apply all bt.
> This actually works out OK, thanks. I had forgotten about core dumps.
> I guess I can add abort() as necessary.
>
>> You can also use the dummy driver (instead of, for example Alsa) with
>> a long --timeout value to be able to step through the code.
> But I wonder what is the problem that's preventing gdb from stopping?
> And why does jack get stuck in a loop printing "subgraph timed out"?
> Even a poorly-behaved client shouldn't crash jack, right?
>
> -- Dan
>
>
> ------------------------------
>
> Message: 4
> Date: Sun, 24 Apr 2011 07:48:49 -0400
> From: Paul Davis<[hidden]>
> To: Dan Muresan<[hidden]>
> Cc: [hidden]
> Subject: Re: [Jack-Devel] gdb-ing jack apps?
> Message-ID:<BANLkTik1nFP3wGo4YKeUFvrk2=[hidden]>
> Content-Type: text/plain; charset=ISO-8859-1
>
> On Sun, Apr 24, 2011 at 7:10 AM, Dan Muresan<[hidden]>  wrote:
>>> One way is to get core dumps from the app (ulimit -c unlimited) and
>>> then gdb ./myapp core, and then in gdb: thread apply all bt.
>> This actually works out OK, thanks. I had forgotten about core dumps.
>> I guess I can add abort() as necessary.
>>
>>> You can also use the dummy driver (instead of, for example Alsa) with
>>> a long --timeout value to be able to step through the code.
>> But I wonder what is the problem that's preventing gdb from stopping?
>> And why does jack get stuck in a loop printing "subgraph timed out"?
>> Even a poorly-behaved client shouldn't crash jack, right?
> i know that some people do seem to have problems using JACK and gdb.
> but others (e.g. myself) never have (modulo the usual issues with not
> running in RT mode and using a long timeout. i routinely run ardour
> inside gdb a lot, and stop in various places without the issue you
> describe.
>
> you didn't mention which version of JACK you are using - the error
> message makes it sound like jack1, but which release?
>
>
> ------------------------------
>
> Message: 5
> Date: Sun, 24 Apr 2011 15:14:40 +0300
> From: Dan Muresan<[hidden]>
> To: Paul Davis<[hidden]>
> Cc: [hidden]
> Subject: Re: [Jack-Devel] gdb-ing jack apps?
> Message-ID:<[hidden]>
> Content-Type: text/plain; charset=ISO-8859-1
>
>> you didn't mention which version of JACK you are using - the error
>> message makes it sound like jack1, but which release?
> Yes, jackd1 -- from Ubuntu Natty (1:0.118+svn4104-1ubuntu2); jackd
> --version reports 0.120.1
>
> http://packages.ubuntu.com/natty/jackd1
>
> -- Dan
>
>
> ------------------------------
>
> Message: 6
> Date: Sun, 24 Apr 2011 23:38:23 +0300
> From: Nedko Arnaudov<[hidden]>
> To: St?phane Letz<[hidden]>
> Cc: [hidden]
> Subject: [Jack-Devel] RFC: fix for missing port registration
> 	notifications	when switching master
> Message-ID:<[hidden]>
> Content-Type: text/plain; charset="us-ascii"
>
> The issue was reported in #213 (second issue/patch):
>
> jackdbus' internal graph representation is not properly kept in sync
> with the engine's graph. In particular: ports are not properly removed
> > From the connection list when a client goes away; and ports
> re-appearing after a "switch-master" are not being updated (Patch
> 2/2).
>
> I think the proposed patch is wrong because AFAIK graph reorder is only
> for ports present upon client activation. All following port
> (un)registrations must cause the port registration callback to be
> called. The proposed changeset [1] adds notification calls in the
> Attach()/Detach() methods. The changeset is commited to a dedicated
> branch [2]. The branch is (hopefully) a short living one with a signle
> commit, so it makes sense to commit it directly to svn trunk (after
> eventually rebasing it, of the svn trunk moved).
>
> Robin, I'd be glad if you can test&  report whether my fix works for
> you.
>
> [1] http://repo.or.cz/w/jack2.git/commitdiff/55557d1f74b6d0ca80981bad9cd646d69c896e0e
> [2] http://repo.or.cz/w/jack2.git/shortlog/refs/heads/switch-master-port-registration-notifications
>


> 7: all i can say is sarcasm is the lowest form of wit. i should know. many thanks,
dave
Prev	Next Index
1303737054.15526_0.ltw:2,a <4DB572C2.1040008 at ntlworld dot com>