On Thu, Feb 19, 2009 at 11:12 PM, Eric Wilhelm <[email protected]> wrote: > With what version of CPANPLUS? I thought the PERL5LIB trouble we were > having came only from the latest CPANPLUS::Dist::Build alpha, but if it > is pre-existing, perhaps we should try a work-around?
The "workaround" is requiring Test::Harness 3.16, which closes http://rt.cpan.org/Public/Bug/Display.html?id=40257 CPANPLUS::Dist::Build just happened to treat PERL5LIB in a way that tickled the bug. It only looked like the problems were with the CPANPLUS::Dist::Build alpha because the alpha finally captures test output and let us see the errors. Once we could see errors, we were able to trace the PERL5LIB problem to T::H and also fix up M::B with allow_mb_mismatch to allow it to test and install itself even under older, broken CP::D::B's. So -- the roadmap is T::H 3.16 -> M::B 0.32 -> CP::D::B 0.07 -> CPANPLUS 0.86 (CPANPLUS 0.86 will automatically bootstrap CP::D::B and M::B if needed and so we want to make sure it bootstraps a new enough version of CP::D::B.) And then it all hopefully goes into blead/maint in time for 5.10.1. And if I find the spare tuits, I'll try to get IPC::Cmd fixed up on Win32 so we can capture test output there, too. -- David
