develooper Front page | perl.perl5.porters | Postings from March 2011

RE: 5.14.0 and MakeMaker (Re: Call for 5.14 release blockers)

Thread Previous | Thread Next
From:
Jan Dubois
Date:
March 25, 2011 10:43
Subject:
RE: 5.14.0 and MakeMaker (Re: Call for 5.14 release blockers)
Message ID:
030101cbeb14$14097af0$3c1c70d0$@activestate.com
On Fri, 25 Mar 2011, Christian Walde wrote:
> On Fri, 25 Mar 2011 15:38:00 +0100, Jesse Vincent <jesse@fsck.com> wrote:
> 
> > If there are specific catastrophic failures in the version of MakeMaker
> > we're shipping that fit the definition of things that would get updated
> > in a maint release, let's talk about them.  I'm happy to buy the beer.
> 
> EUMM 6.57_05 has compilation issues under ActivePerl as shown here:
> http://www.cpantesters.org/cpan/report/f3e82aaa-6c10-1014-a463-7ea05919af89

I do get the t/miniperl.t failure, but not the t/xs.t one.  I do get
an additional t/MM_Win32.t failure though, which looks bogus to me,
but is also still present in 6.57_07.  I assume the difference in
failures is that I built using VC (and namke), not GCC (and dmake).

However, I don't get any failures while building blead (and just 5 warnings
about "uninitialized value in lc" from MM_Any.pm.

> and also with true.pm and perl5i. (It is unable to find gcc or deal
> with .a/.lib correctly.) These issues are present with many CPAN
> modules and have only recently been adequately analyzed and resolved
> with considerable amounts of testing added in 6.57_07 so its
> compatibility with XS distributions on the Win32 is now massively
> increased. (Once a new release is done.)

I believe this is a different issue that only happens when:

1) The user is using GCC to build a module for use with a Perl compiled
   by VC (using MinGW to build modules for ActivePerl) *and*

2) The module is linking against the library of a different module *and*

3) The "different" module has been compiled with VC.

So it affects modules like true.pm, which uses ExtUtils::Depends to
reference functions inside B::Hooks::OP::Check. But only when true is
being compiled with GCC while B::Hooks::OP::Check was compiled with VC
(e.g. because it was installed via PPM from a repository that was
built with VC).

Given that ActivePerl still contains some local patches to MakeMaker, I
think I would prefer to keep blead as-is, and apply another local
change to make it work for the ExtUtils::Depends stuff while mixing
compiler versions.
 
ActivePerl is the only distribution that supports using a different
compiler for building XS code AFAIK, so I don't think this issue affects
any other users.

And yes, I was hoping to get the local ActivePerl changes folded back
into the CPAN version, but didn't find the time (the hard/ugly part
is the additional support for generating HTML docs, that was already
in the CPAN version once, but has been ripped out again...).

> (As an aside: Most of the delay was my fault because i wanted to give
> people a chance to look over my changes.)

I'm sorry, you were probably waiting for me there. :(

Cheers,
-Jan


Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About