jack for lenny

[Reinhard Tartler]

Loïc Minier <lool at dooz.org> writes:

>> thread.h changes the binary structure of jack_position_t.
>
>  Is it an internal only structure, or is it exposed to applications to
>  pass around jack's ABI?

Well, it is in an public header, so I suspect it to be used by
applications. I've therefore downloaded all packages build-depending on
libjack-dev (I used the versions in lenny) and identified the following
packages referencing that type (with a simple grep, false positives are
possible!):

mscore
muse
openmovieeditor
qtractor
rosegarden
traverso

>  Are applications doing things like
>  jack_position_t * ppos = jack_position_new(); jack_position_set(ppos)
>  etc., or are applications doing things like
>  jack_position_t pos; pos.foo = bar
>  ?

No idea, we would need to investigate the source code of the packages
listed above. But I'm not sure if that really matters, because the
compiler of applications that have been built against 0.109 would almost
certainly have a different idea of the binary layout compared to the
structure in version 0.116, if there is really a problem.

Free, does your testing include the above packages from lenny (read:
packages that have been built with version 0.109) with jack 0.116?

And moreover, do you have some more numbers about the testcoverage?
Like how the testing was done and how many users have tested the
relevant scenario (read: packages built against 0.109 and being used
with libjack 0.116)?

TBH, from the source code inspection, I do expect problems. We could of
course go ahead and patch out that change, and recompile. If we split
out the libjackserver.so library to some new pacakge libjackserver0, we
can *perhaps* maintain shlibs. But I'm not yet convinced that this would
be worth the efford. :-(

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4