static libs ... again

[Hans de Goede]

Rex Dieter wrote:
> Hans de Goede wrote:
>> Rex Dieter wrote:
> 
>>> Eek.    I still think headers and api docs and such still should be 
>>> in -devel (especially if there's any likelyhood of a real shared lib 
>>> existing some day), and that -static should Requires: %{name}-devel
> 
>>>> Also I wonder how hard is it to add -fpic -DPIC to the cflags and 
>>>> change the link command to generate an .so. The only added trouble 
>>>> would be checking for abi changes on new releases and bumping the 
>>>> .so name a release.
> 
>>> Exactly.  I'm of the opinion (in most cases) that if upstream isn't 
>>> able/willing to do something (like generating shared libs), then 
>>> neither am I (as packager).
> 
>> You say "Exactly" as in I agree with you and then you continue with 
>> saying that you're not willing todo this, I'm confused now.
> 
> Exactly, as in "The only added trouble..." part.  (-:
> 
> As I said, if it's really not so hard, let upstream do it, per Fedora's 
> mantra "Upstream, upstream, upstream..."
> 

Yes,

But sometimes upstream doesn't, because they dont care about this for 
example, yet it would still be worth the trouble. I believe this needs 
more discussion we all seem to agree that static libs should be provided 
if possible, since in most cases even if upstream doesn't do it it isn't 
all that much work, why don't we do this.

I'm planning on packaging some software that uses one of these only has 
-devel with static libs packages with no .so and I'm actually also 
planning on filing an RFE against this package for proper .so files.

Why shouldn't a packager do this? Upstream is what we want, but 
unfortunatly is not always what we get.


Regards,

Hans