Adrian Knoth wrote:
> On Sat, Dec 19, 2009 at 05:02:21PM +0100, hungerburg wrote:
> 
> Hi!
> 
>> the last time I used ardour two days ago. I believe it was
>> 1:2.8.4-1 - today after updating to 1:2.8.4-2 I can no longer
>> launch the application. A downgrade does not help. Maybe its
>> because a file changed that ardour uses but that is in another
>> package.
> 
> We have to sort things out. I claim that it's not an ardour bug,
> ardour's working fine over here and you say that downgrading doesn't fix
> the bug, so it's very likely not ardour's fault.
> 
> I'll reassign the bug to the appropriate package once we know the
> culprit.
> 
>> Below valgrind output shows an invalid memory access while
>> loading lv2 plugins. I am not aware that debian ships any of
>> those though or that such plugins are installed on this
>> system here.
> 
> We do have LV2 plugins in Debian, however, ardour's LV2 code is rather
> stable. See below.
> 
>> ==29391== Jump to the invalid address stated on the next line
>> ==29391==    at 0xC: ???
>> ==29391==    by 0x58CEF8F: rasqal_engine_execute_init (in 
>> /usr/lib/librasqal.so.1.0.0)
>> ==29391==    by 0x58CA128: rasqal_query_execute (in 
>> /usr/lib/librasqal.so.1.0.0)
>> ==29391==    by 0x58A8781: ??? (in /usr/lib/librdf.so.0.0.0)
>> ==29391==    by 0x58A5D49: librdf_query_execute (in /usr/lib/librdf.so.0.0.0)
>> ==29391==    by 0x595A83D: slv2_world_load_specifications (in 
>> /usr/lib/libslv2.so.9.2.0)
>> ==29391==    by 0x595AE93: slv2_world_load_all (in /usr/lib/libslv2.so.9.2.0)
>> ==29391==    by 0x424E262: ARDOUR::LV2World::LV2World() (in 
>> /usr/lib/ardour2/libardour.so)
>> ==29391==    by 0x4198453: ARDOUR::PluginManager::PluginManager() (in 
>> /usr/lib/ardour2/libardour.so)
>> ==29391==    by 0x41462D0: ARDOUR::init(bool, bool) (in 
>> /usr/lib/ardour2/libardour.so)
>> ==29391==    by 0x816080C: ARDOUR_UI::ARDOUR_UI(int*, char***) (in 
>> /usr/lib/ardour2/ardour-2.8.4)
>> ==29391==  Address 0xc is not stack'd, malloc'd or (recently) free'd
> 
> Here we go. As always, stack traces are bottom up. We see that ardour
> calls some LV2 functions which are handled by libslv2. This library does
> the whole LV2 discovery stuff.
> 
> Further up the stack trace, we see librdf executing a query, which is
> then passed to librasqal.
> 
> Both libs, librdf and librasqal, have new versions in unstable since
> yesterday, so I guess one of them is the culprit:
> 
>    http://packages.qa.debian.org/r/redland.html
> 
>    http://packages.qa.debian.org/r/rasqal.html
> 
> (new versions on 2009-12-17)
> 
> 
> When I downgrade librdf0 to 1.0.9-3, ardour starts fine.
> 
> 
> That's why I would say librdf0 is the real culprit here.
> 
> I'm now building ardour against new librdf0 to see if this fixes the
> issue, but I don't think so. (libs are expected to be backward
> compatible, that is, ardour built against librdf0-1.0.9 should run with
> librdf0-1.0.10.
> 
> 
> However, I'll CC the librdf0 maintainer, so he has the chance to do some
> investigations himself.
> 

>From the stack trace I can see a problem - the librdf is calling the old
rasqal ABI (0.9.16) in librasqal.so.1 rather than the new (0.9.17) in
librasqal.so.2.  (Assuming these are redland librdf 1.0.10 and rasqal librdf
0.9.17)

redland 1.0.10 (librdf.so.0) was packaged to link with the new rasqal 0.9.17
(librasqal.so.2) and so the above stack trace should not be possible, unless
I've got the shared lib/package dependencies wrong.  Maybe librasqal2 should
conflict with librasqal1.

Dave



_______________________________________________
pkg-multimedia-maintainers mailing list
pkg-multimedia-maintainers@lists.alioth.debian.org
http://lists.alioth.debian.org/mailman/listinfo/pkg-multimedia-maintainers

Reply via email to