On Tue, May 31, 2016 at 1:28 PM, Craig A. Berry <craig.a.berry@gmail.com> wrote: > On Tue, May 31, 2016 at 12:40 PM, Craig A. Berry > <craig.a.berry@gmail.com> wrote: > >> # pl2bat.bat blib\script\program >> # C:\Perl64\bin\pl2bat.bat: Perl lib version (5.25.2) doesn't match >> executable 'C:\Perl64\bin\perl.exe' version (5.20.2) at >> C:\perlgit\lib/Config.pm line 62. >> # Compilation failed in require at C:\Perl64\bin\pl2bat.bat line 20. >> # BEGIN failed--compilation aborted at C:\Perl64\bin\pl2bat.bat line 20. > > The FIXIN macro in EUMM, when PERL_CORE is not set, just runs the > installed pl2bat.bat. See: > > <https://github.com/Perl-Toolchain-Gang/ExtUtils-MakeMaker/blob/2df481c3227628992c5ad92a2ecb7d7ddb0a090a/lib/ExtUtils/MM_Win32.pm#L145> > > This recent commit: > > <https://github.com/Perl-Toolchain-Gang/ExtUtils-MakeMaker/commit/4bd05555380a30ea7e532bbe8884607b874c420e> > > deletes PERL_CORE from the environment in basic.t. I have no idea why > and the commit message doesn't say. But I think that's the > explanation for why it's running the installed pl2bat.bat in my case, > and probably fails differently if there is no pl2bat.bat in the path > (e.g., there is no installed Perl). The problem behind this problem is that when PERL_CORE is set, running Makefile.PL does not generate the metafile section of the Makefile. That happened last September: <https://github.com/Perl-Toolchain-Gang/ExtUtils-MakeMaker/commit/f91b5144ee197e0c875e927e5b2603d65d0797d1> I'm guessing there was a desire to test metafile and other things depending on it (e.g., disttest) even when running in core, thus the deletion of PERL_CORE in some tests more recently. It looks like that only causes trouble on Windows because of the FIXIN problem cited earlier. Not sure what to suggest. It's two different meanings of PERL_CORE colliding: 1.) "do things differently when using MakeMaker to build the core" and 2.) "do things differently when running MakeMaker's own tests in the core."Thread Previous | Thread Next