Bug#565342: newly released libffado2-based jackd fails with "firewire ERR: Could not start streaming threads: -1"

[Adrian Knoth]

On Sat, Jan 16, 2010 at 05:20:03PM +0100, Jonas Smedegaard wrote:

Hi!

>> If need be, file a ticket at subversion.ffado.org
> I guess it makes sense to first try isolate which part of the Debian  
> packaging system cause problem before passing it upstream:

That's a good idea. Thanks for your comprehensive tests.

> Results using *jack* 0.118+svn3796-1, *ffado* 2.0~rc2+svn1569-2 and  
> *freebob* 1.0.11-1 (i.e. latest Debian Sid except jackd and least  
> possible dependencies rolled back to latest Debian Squeeze):
>
> OK: firewire-ohci + jackd -d firewire -p 512 -n3 -r 44100 -v5
> OK: dv1394 + raw1394 + jackd -d firewire -p 512 -n3 -r 44100 -v5
> OK: dv1394 + raw1394 + jackd -d freebob -p 512 -n3 -r 44100

So to sum this up: ffado-2.0~rc2+svn1569-2 gives good results. Freebob
doesn't matter here, and I'd for now say that jackd version doesn't
matter, but we'll soon find out...

> Above makes me assume that hardware (MacBook) is ok.

I would say so, yes.

> Results using *jack* 0.118+svn3796-2, *ffado* 2.0.0-1 and no *freebob*  
>
> OK: dv1394 + raw1394 + jackd -d firewire -p 512 -n3 -r 44100 -v5
> FAIL: firewire-ohci + jackd -d firewire -p 512 -n3 -r 44100 -v5

That's interesting. It would indicate that ffado itself is "fine", and
it's somehow a question between ffado and the new stack in this
particular version.

There has been a small update to FFADO to make it run on Juju stacks,
however, your tests with the rc2+svn1569 indicate that it has been
(somehow) working before.

I checked

   http://subversion.ffado.org/ticket/240

and it looks like the FFADO-on-Juju fixes never got applied to the 2.0.0
release branch.


Any chance you can rebuild libffado from source and apply a patch,
first?

I'm thinking of:

   # apt-get source libffado
   # apt-get build-dep libffado 
   $ cd libffado-*
   $ patch -p1 < /path/to/ffado-2.0.0.patch
   $ debuild
   $ sudo debi


I've attached ffado-2.0.0.patch to this mail. I guess this could make
ffado work on Juju. If so, I'll take care that this patch gets applied
upstream and that we get another release, perhaps ffado-2.0.1.

> So, it seems to be that problem is in either jack compilation/linkage or  
> in FFADO source.

I bet it's FFADO source, but we'll find out:

> How to proceed from here?  More tests I can perform?

You can test ffado-2.0rc2+svn1569-2 (the testing version) on new jackd.
Simply install libffado1 and overwrite/symlink to libffado2 in /usr/lib.

This way, both, old and new jackd versions can be tricked to pick up
either libffado1 or libffado2. (these libs are API/ABI compatible at the
moment, so this dirty hack can work.)

This makes sure jackd or linkage isn't the culprit.


> Should I post above to upstream FFADO or Jackd developers (and if so,
> how?)?

Let's first try the attached patch. It has nothing to do with the jackd
developers, since jackd code hasn't changed, only ffado.

I've already pinged ffado upstream on this bug report, and I'm somewhat
ffado upstream myself, so I'll take care of everything.



Cheerio

-- 
mail: adi at thur.de  	http://adi.thur.de	PGP/GPG: key via keyserver
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ffado-2.0.0.patch
Type: text/x-diff
Size: 3813 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-multimedia-maintainers/attachments/20100116/ef85bcaf/attachment.patch>