Re: when can a package be made architecture-dependent?
"Steven G. Johnson" <stevenj@ab-initio.mit.edu> writes:
> On 18 Jan 2003, Adam DiCarlo wrote:
> > But I do think this goes too far. There might be good reasons why the
> > upstream maintainers or debian maintainers are unable to maintain a
> > ported package -- notably, if the upstream were not willing to take
> > patches for building in other architectures.
>
> In which case the build will fail on those architectures, but as was
> pointed out earlier this will not keep the package out of testing/stable
> if the package never built on those architectures, so what's the problem?
If neither the upstream maintainer(s) nor the debian developer is
willing to maintain the port, you are proposing that the Debian
developer ought to maintain the port. I think this is wrong in
principle and wrong in fact. I think if the Debian pkg maintainer
does want to maintain the port -- fine. If the package is really
inherently portable, fine. But in the other case, it's really the
maintainer's call to support the arch or not, I think.
Take the example of a piece of software where certain core routines
need to be written in assembler for the software to work. Assembly
routines are only written for X86 and SUN4 CPUs. The Debian pkg
should be available only for i386 and sparc in this case.
Take the example of Java. Only certain arches are supported and
provided by Blackdown. I think it makes sense if the Debian packages
reflected that.
Doing anything else is deliberately forcing buggy packages onto the
overworked porters.
> Labelling it correctly will make it easier to identify which packages
> could use porting,
If all they want is a list of packages in i386 and not in some other
arch, there are better ways to get it.
> keep people aware of build bugs, prod the packagers to
> make at least a minimal effort at maintaining portability, etc., and
> thereby make the porters' job easier.
Sure, in the case you cited, it was just a build bug, fine, I agree.
I'm just saying (again) that I thought your expansion went a little
too far. I don't wnat to encourage Debian forking.
> But I'll leave this for the Debian policy wonks to debate.
Well, thankfully we're out of the realm of policy and into the realm
of "best practices".
--
...Adam Di Carlo..<adam@onshore-devel.com>...<URL:http://www.onshored.com/>
Reply to: