Linux Archive

Linux Archive (http://www.linux-archive.org/)
-   Debian Development (http://www.linux-archive.org/debian-development/)
-   -   Multiarch and D-Bus (http://www.linux-archive.org/debian-development/546830-multiarch-d-bus.html)

Aurelien Jarno 07-01-2011 08:54 AM

Multiarch and D-Bus
 
On Fri, Jul 01, 2011 at 09:23:33AM +0100, Simon McVittie wrote:
> On Fri, 01 Jul 2011 at 10:06:34 +0200, Bastien ROUCARIES wrote:
> > Even dark corner like double endianess ?
>
> Doubles are byteswapped in exactly the same way as int64, I hope that's
> correct everywhere? (dbus/dbus-marshal-byteswap.c around line 73.)
>
> To be more specific: D-Bus assumes that doubles are IEEE-754, in the same
> byte order as integers - so all LE architectures have the same double format
> as an i386, and all BE architectures have the same double format
> as a powerpc. Do we have any architectures where this isn't true?

We used to have the ARM old-ABI architecture using mixed-endianness, but
we don't have this architecture anymore (replaced by armel).

While the float and doubles are going to be transfered correctly for all
Debian architectures from the endianness point of view, architectures
will interpret some special values differently depending on the
architecture. For example depending on the architecture, a qNaN and sNaN
values can be swapped.

I don't know if this is handled correctly in D-BUS.

--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110701085412.GI11516@hall.aurel32.net">http://lists.debian.org/20110701085412.GI11516@hall.aurel32.net

Simon McVittie 07-01-2011 09:07 AM

Multiarch and D-Bus
 
On Fri, 01 Jul 2011 at 10:54:12 +0200, Aurelien Jarno wrote:
> We used to have the ARM old-ABI architecture using mixed-endianness, but
> we don't have this architecture anymore (replaced by armel).

D-Bus interop for doubles was always broken on ARM old-ABI, then, and nobody
noticed :-(

The current D-Bus Specification only has two possible endianness flags,
'l' and 'B'; adding proper support for mixed-endianness would be an
incompatible change, unfortunately. I suspect architectures not supported
by Debian are too obscure to worry about, though.

> While the float and doubles are going to be transfered correctly for all
> Debian architectures from the endianness point of view, architectures
> will interpret some special values differently depending on the
> architecture. For example depending on the architecture, a qNaN and sNaN
> values can be swapped.

Thanks, I'll bring this up upstream, although I suspect the response will be
"don't send NaN, then". We don't have any special handling for NaN or other
unusual values at the moment.

S


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110701090725.GA19279@reptile.pseudorandom.co.uk" >http://lists.debian.org/20110701090725.GA19279@reptile.pseudorandom.co.uk

Samuel Thibault 07-01-2011 09:49 AM

Multiarch and D-Bus
 
Josselin Mouette, le Fri 01 Jul 2011 09:39:54 +0200, a écrit :
> but if we can get rid of at-spi and
> have GConf ported to D-Bus before the wheezy release, I’m not sure it’s
> worth an investigation.

For now, at-spi2 is very far from being as usable as at-spi, it would be
a major accessibility regression. I however believe corba can be used
over TCP/IP, and thus probably has arch-neutral support.

Samuel


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110701094910.GF4279@const.bordeaux.inria.fr">htt p://lists.debian.org/20110701094910.GF4279@const.bordeaux.inria.fr

Wouter Verhelst 07-01-2011 11:49 AM

Multiarch and D-Bus
 
On Fri, Jul 01, 2011 at 11:49:11AM +0200, Samuel Thibault wrote:
> a major accessibility regression. I however believe corba can be used
> over TCP/IP,

I would be *very* surprised if that wasn't true -- CORBA was meant to be
used over TCP/IP, using it to localhost or over a unix domain socket is
just a lucky accident.

--
The volume of a pizza of thickness a and radius z can be described by
the following formula:

pi zz a


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110701114928.GB5420@grep.be">http://lists.debian.org/20110701114928.GB5420@grep.be

Wouter Verhelst 07-01-2011 11:49 AM

Multiarch and D-Bus
 
On Fri, Jul 01, 2011 at 11:49:11AM +0200, Samuel Thibault wrote:
> a major accessibility regression. I however believe corba can be used
> over TCP/IP,

I would be *very* surprised if that wasn't true -- CORBA was meant to be
used over TCP/IP, using it to localhost or over a unix domain socket is
just a lucky accident.

--
The volume of a pizza of thickness a and radius z can be described by
the following formula:

pi zz a


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110701114928.GB5420@grep.be">http://lists.debian.org/20110701114928.GB5420@grep.be


All times are GMT. The time now is 07:00 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.