random comments about secondary arch proposal

[Thorsten Leemhuis]

On 15.06.2007 18:04, Axel Thimm wrote:
> On Fri, Jun 15, 2007 at 11:43:07AM +0200, Thorsten Leemhuis wrote:
>>
>> On 14.06.2007 19:40, Christopher Blizzard wrote:
>>
>>> Primary arch definition: need to make sure that part of the
>>> responsibility is that individual maintainers are required to make sure
>>> their stuff builds on all those arches.
>> -1 -- Fedora Extras had lots of maintainers that are no programmers
>> and/or have only access to i386.
> Don't they have access to *build* their package through koji by
> definition on all archs?

That's a help only for "have only access to i386" people; and even then
it's often a lot easier to do it on a machine you have direct access to
then to

until builds_on_ppc()
do
	look at the code
	write a patch to fix a issue
	commit a patch to CVS
	tag
	build
	wait
done

>> Those maintainers are Fedora maintainers these days. Saying they are
>> "required to make sure their stuff builds on all primary arches" would
>> increase the burden on the maintainer drastically. I think that's
>> totally the wrong way forward.
>>
>> Further: if I would read something like that before becoming a
>> contributor I'd say "hey, that's hard stuff; I know my knowledge is not
>> enough to do that should I even run into a situation where something
>> doesn't build on PPC; well, then I won't become a contributor for
>> Fedora. Have fun guys, bye".
> But pcc is not a primary arch, or is it?

There is not final decision yet (or I missed it), but afaics it looks
like PPC will be one  at least for the near future until the secondary
arch stuff works fine for another arch.

> [...]
> I would go as far as to suggest test boxes on i386/x86_64
> (e.g. primary archs) for packages to test their koji builds on in case
> they want to, but don't have the archs available otherwise. Sure, some
> packages like the kernel and glibc can't just be tested on a rmeote
> machine, but I hope the maintainers of these packages have access to
> more than just an i386 system ;=)

Yes, something like that would be nice to have; but I'm not sure if
other stuff would be more important (automatic tests for example).

CU
thl