RFC: Better portability for package maintainers

[Alex Ross]

Matt Zimmerman wrote:
> On Fri, May 19, 2006 at 01:15:44PM -0700, Alex Ross wrote:
>> Matt Zimmerman wrote:
>>> On Fri, May 19, 2006 at 10:34:35AM -0700, Alex Ross wrote:
>>>> The following is based on premises that portability is good and that
>>>> POSIX is a standard. A proposal.
>>> I didn't see a concrete proposal in your email, only information about
>>> where to find gnusolaris build logs.  Can you elaborate?
>>>
>> Minimally, package maintainers and developers could take a look on our logs
>> and see if there's anything wrong. If there is, in many cases the fix is 
>> obvious.
> 
> Because Debian and Ubuntu developers may not have access to a similar build
> environment, it will be difficult for them to help you in this way.  Even if
> the first error is fixed, they will need to wait for you to build the
> package again to see if it builds.  I expect that most maintainers won't be
> interested in working this way.  

Ditto.

> It makes more sense for the Nexenta
> community to fix these problems and make the patches available to other
> distributions and upstream.  If your intention is to invite the Debian and
> Ubuntu developer communities to participate in this way, perhaps a good step
> would be to offer them access to a porting machine where they can help you
> with this work if they are interested?

ssh?.. We'll work on it, let you know.

> 
>> Ideally though, there'd be an augmented policy of package acceptance,
>> reflecting the fact that the packages with "Architecture: any" should build
>> and run on one of the Debian POSIX-compliant systems. NexentaOS is
>> certainly one such system. To help implement this new policy we could "plug"
>> our AutoBuilder [1] into the existing build environment.
> 
> If we find that it is feasible to get the current packages in Debian
> to meet this standard, then we can consider extending policy to recommend or
> mandate it at some point in the future, but to do this before attempting the
> experiment would be premature.
>