Re: INSTALLPRIVLIB should be /usr/lib/perl5

[Roderick Schertler <roderick@argon.org>]

On Wed, 14 Jul 1999 22:05:25 +0200, Raphael Hertzog <rhertzog@hrnet.fr> said:
> Le Wed, Jul 14, 1999 at 02:36:34PM -0400, Roderick Schertler écrivait:
>>
>> Both the 5.005 and 5.004 builds have $Config{installprivlib} set to
>> /usr/lib/perl5/$version.  I think this is a mistake, the distributed
>
> No, it's the upstream way of doing things. And we follow for the most
> part the upstream perl ...

That doesn't make sense.  The upstream version doesn't keep a separate
directory for Perl-only non-site_perl modules the way we do.

> It saves you one line, and if you're really lazy, you can even cut and
> paste it from the perl policy.

The problem isn't the amount of work it takes to add one line to a rules
file, it's that there's one more thing which has to be done for every
module, and one more thing which can be done wrong.  There's a choice
between making this change once (when the Perl package is built) and
making it dozens of times (when the module packages are built).

I think it'd be useful to take this a step further.  Right now at a
minimum a rules file has to say

    $(MAKE) \
	    PREFIX=$(prefix)/usr \
	    INSTALLPRIVLIB=$(prefix)/usr/lib/perl5 \
	    INSTALLMAN1DIR=$(prefix)/usr/man/man1 \
	    INSTALLMAN3DIR=$(prefix)/usr/man/man3 \
	pure_perl_install

in order to do the right thing with all modules.  That's too much
boilerplate.  I suggest it would be useful to have a way to, say, pass
an extra arg on the Makefile.PL line which would take care of most of
this, perhaps leaving only the prefix to bet set in the rules file.

In a similar vein, the Perl policy suggests you compile with

    make CFLAGS="-O2 -g -Wall"

That should also not be necessary, the defaults should be whatever is
correct.

> And last but not least, the installarchlib is calculated with
> installprivlib when not overriden on the command line ... that
> would mean that if we use installprivlib then we still need to add
> a line for the binary modules.

I don't follow you here.  Could you elaborate?  Where is this calculation
happening?

-- 
Roderick Schertler
roderick@argon.org