Re: [Chicken-hackers] pkg-config support in chicken-install

[Lassi Kortela]

> I can't comment on that, I don't know enough about it, with the
> exception of pkg-config related errors I darkly remmeber, but
> that may have been caused by my borked setup.

It can get borked, but the key thing is that we have someone responsible 
(the packager who supplied the pkg-config definitions) and an actionable 
plan for how to fix things (ask them to fix their definitions).

If we write custom configure-style scripts and they break, we don't have 
any clout to blame anyone but ourselves, or any standard procedure that 
the upstream packager can follow to fix things for us.

> > it'd be very useful if the egg description lists the names of the
> > requisite OS packages.
>
> Yes, but these names aren't even standardized across Linuxen, as
> I've been told. A truly reliable approach sounds very much effort to
> me. But, ok, one can at least try...

They are not standardized, but the situation is not as bad as it seems 
:) There are not that many alternative names that distributions use for 
a given upstream package. And our lists don't have to be complete - a 
partial list is still very useful in helping users find the packages. 
Assume that for example 50% of users have Debian or Ubuntu. Then, even 
if we specify only the Debian APT package names and nothing else, we can 
already help all of those people. And the more package names we add the 
more people we help.

We don't have to promise that "if you install this package from the OS 
package manager, your Chicken egg is 100% guaranteed to work". We can 
just say that "this will probably work on your operating system". If we 
don't have any package name for a particular OS, we can simply be silent 
(which is the situation we have now, so it's no worse). As a user, I 
appreciate such a time-saver a lot - if I don't get that advice, then 
I'll easily spend 30-60 minutes googling for help, and there's no 
guarantee that the solutions from Google results work either... So IMHO 
offering this imperfect help is a useful improvement for a modest cost.

> I'm not completely opposed but extremely wary. And nevermind who's
> implementation it is (I couldn't say) and there are no outsiders. Seeing
> a prototype would be actually pretty helpful, since I have no clear idea
> how this should be implemented.

That's very modest and welcoming of you :)

Thanks for considering it. I'll copy the S-expression metadata for the 
openssl egg and add OS package and pkg-config information to it. Then 
I'll write a mock-up command line program in Chicken 5 that shows what 
chicken-install would do given that file (i.e. calling pkg-config and 
showing messages if it can't find the pkg-config program or the C 
libraries).