>>>>> On Tue, 10 Apr 2007 10:38:04 -0700, "Joshua ben Jore" <[EMAIL PROTECTED]> >>>>> said:
>> I could imagine you didn't use CPAN.pm 1.90, right? > For sure. I was testing that a "fresh" 5.8.8 could install it without > error. I use this to help suss out undeclared prerequisites. I have > the same environment for all of 5.6.2/5.8.1->5.8.8/5.9.5 under > non-/threaded, non-/debugging. While I did it manually today, I've got > a harness over it all so I can essentially just request that *all* > those perls attempt to go install a CPAN module of my choice (or a > local distribution) and I get TAP back about which perls passed and > failed. > # for a single perl and what happened last night > cd /opt > rm -rf perl-5.8.8/* > tar xjf perl-5.8.8.tar.bz2 # a fresh 5.8.8 but with updated CPAN::Config > /opt/perl-5.8.8/bin/cpan Catalyst < /dev/null I'm sorry, I have only two possible explanations: either there was a bug in an older version of CPAN.pm that I absolutely cannot recall. I cannot imagine that this is ever possible to happen with current codebase. Or there is one of these Module::Install gremlins at work and thwarts otherwise solid code by calling CPAN.pm recursively. All bets are off when this is happening. See https://rt.cpan.org/Ticket/Display.html?id=23735 for my warnocked bugreport. 1.90 now desparately sets the environment variable PERL5_CPANPLUS_IS_RUNNING (!) to alleviate the effects of this bug in Module::Install. You can probably also set PERL5_CPANPLUS_IS_RUNNING yourself for older versions of CPAN.pm. Maybe it helps. Or if you have the output of that session, please send it to me in private mail, I'll try to find out what happened. -- andreas
