[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#746588: gnat should be an architecture dependent package

Ludovic Brenta dixit:

>It is our policy[1] to support the same version of gnat on all
>architectures.  The package gnat exists for this very purpose.  Allowing

Right, but there is still the transition period to take care of.

>gnat to depend on different versions of gnat-x.y on different
>architectures would defeat this purpose.  If some architecture cannot

This is not the intention. The idea here is that, for as long
as the _new_ version of gnat cannot _yet_ be built, the old
version can stick around for some architecture. We will *see*
the new version, and know there is something to do, but until
such time, we can still use the old version.

This is *especially* important because gnat-x.y Build-Depends
on gnat, which means we *must* keep a working gnat metapackage
in the archive.

>support the chosen version of gnat-x.y, then we prefer to drop support
>for that achitecture altogether.

Really? Especially debian-ports.org architectures can sometimes
be a bit slower in adopting things like this, but gnat *can*
work there. Do you really want to throw this into porters' faces?

I've been spearheading porting gnat to m68k. Not because I use
it, but as a favour to Ada users, and for Debian completeness.
Not quite the return attitude I expected... especially for all
the pain involved (cross-compiling, several times rebootstrapping,
writing patches myself, getting patches from the GCC people,
testing stuff, learning about the m68k arch more, etc).

>The transition from gnat-4.6 to gnat-4.9 has been planned and announced
>in due course [3,4] (and furthermore accepted and approved by the Ada

Yes, but not on debian-ports@l.d.o - and besides, that's not the
point. We'll follow your transition, as we do with everything that
is uploaded to unstable, but on a timeline that works with the ports.

>objections, please explain the benefits of making it easier to break

There is no break of any policy involved. If something depends
on gnat 4.9, it must have 'gnat (>= 4.9)' in Build-Depends anyway,
for at least until gnat 4.9 (distinguished from gnat-4.9) is in
stable or even oldstable, which is true for all versioned Build-Depends
in Debian.

“It is inappropriate to require that a time represented as
 seconds since the Epoch precisely represent the number of
 seconds between the referenced time and the Epoch.”
	-- IEEE Std 1003.1b-1993 (POSIX) Section B.2.2.2

Reply to: