On 4/17/07, Eric Wilhelm <[EMAIL PROTECTED]> wrote:
# from Paul LeoNerd Evans
# on Tuesday 17 April 2007 02:14 am:
>I ran into similar issues; I've discussed them earlier on this list.
Yep.
# from David Golden on Sunday 11 March 2007 05:43 am:
>This bug is the result of CPANPLUS trying to be clever and "falling
>back" to generating a Makefile.PL from a Build.PL.
My thinking was that exposing the bug where it lives is probably a
better long-term approach.
>The suggestion people made was to have M::B produce a
> 'traditional'-style makefile as well. So at least M::B-deprived
> systems might have a good go.
My problem with the create_makefile_pl => 'traditional' is that it
doesn't install Module::Build, so we're not making any progress. If
you have a custom build subclass, you might as well put a die in
Makefile.PL.
The passthrough seems much more sane because it actually installs
Module::Build, yet nobody seems to encourage its use. Why?
The passthrough, IIRC, usually broke stuff. It might be better now
that M::B has had another year of bugs shaking out.
The problem with having a Makefile.PL that dies is now you've
presented an interface that can't be queried against. Consider this
like duck typing - your distribution ->can( 'Makefile' ) but when you
call the method it dies. That's reasonable if your distribution can
actually use a Makefile. When it can't, its just making the job of
toolchains harder.
At least with an absent Makefile.PL, its possible to look at a
distribution and sanely conclude that there's no Makefile.PL. Or do
you wish for a protocol for fake Makefile.PL files to communicate to
toolchains that "Ooops! No, don't use me!"?
This comes because you want nicer data from CPAN smokers, right? I
already get dumb failures where the smoker registered the fact that my
modules don't pass tests when the prerequisites aren't installed.
"Amazing." That's better fixed with a note to the person running the
smoker than a fix to the CPAN distribution.
Josh
