Re: [Jack-Devel] JACK 1.9.10 to test (for 64/32 bits compatibility)
Robin Gareus <[hidden]> writes:
>> On Wed, Jan 09, 2013 at 10:54:12AM +0100, Stéphane Letz wrote:
>>
>>> I stay on my latest view: do not try to solve a problem (possible
>>> 64/32 Bits issue on ARM) that is still not there.
But the problem with ABI mismatches between jack1 and jack2 is not
solved. Or rather jack2 and any other jack implementation that
uses the non-packed structs in the API headers. Or I am missing
something?
>> [0] From http://wiki.debian.org/ArmEabiPort#GCC_view
>>
>
> [0] on /Struct packing and alignment/:
> "With the new ABI, default structure packing changes [..]. This will
> break programs that know too much about the way structures are packed.."
Thus the current "force alignment" approach is not good. Because
alignment (at least in gcc) specifies minimum alignment, not enforced
one. You can request 8 but get 16.
JACK code should either
use natural alignment (all components are built with same ABI) and get
all but 32/64 bit (wine) problems solved
or
use the packed structs with explicit padding approach and get all
problems solved with more work by JACK developers.
--
Nedko Arnaudov <GnuPG KeyID: 5D1B58ED>
1357743518.10091_0.ltw:2, <87fw2ao12p.fsf at arnaudov dot name>