The below patch reduces quite significantly the time "make test" takes on my system for distributions with around 50 files in lib/... because firing up perl -MExtUtils::MM_Unix -e 1 takes 0.2 to 0.3 seconds and doing it 8 times for each make test (from one dist I work with frequently) adds up. The patch cust make test from 5 seconds to less than 2. Sounds like a small increase, but I run tests very frequenty (via Test::Verbose's tv command). I'm not sure what an acceptable solution is here, but limiting line length to 200 characters on all systems seems overly conservative and detunes the testing mechanism enough to gently deter people from running tests frequently. Limiting line length to 20000 is just an easy number for demonstration purposes; personally, I liked 5.6.1's behavior of unlimited line length better, but there is some magic number for each system and it's very, very big (relative to 200 :) on Linux. How can a real number for line length be determined, or can we avoid line length entirely and pass the hash in through STDIN? If it can't be, I'll submit a patch to MM* to let the user determine this so testing wonks like myself can blow out the limits where appropriate ;). Also, why not have MM_*.pm generate a ruleset like: pm_to_blib: $(TO_INST_PM) blib/lib/Foo.pm @$(TOUCH) $@ blib/lib/Foo.pm: lib/Foo.pm @$(PERLRUNINST) "-MExtUtils::Install" \ -e "pm_to_blib({qw{lib/Foo.pm blib/lib/Foo.pm}},'$(INST_LIB)/auto','$(PM_FILTER)')" and let make(1) do what make(1) is supposed to do? - Barrie --- MM_Unix.pm.orig 2003-02-27 04:07:57.000000000 -0500 +++ MM_Unix.pm 2003-02-27 04:08:02.000000000 -0500 @@ -3110,7 +3110,7 @@ while (my ($pm, $blib) = each %pm_to_blib) { my $la = length $pm; my $lb = length $blib; - if ($l + $la + $lb + @a / 2 > 200) { # limit line length + if ($l + $la + $lb + @a / 2 > 20000) { # limit line length _pm_to_blib_flush($self, $autodir, \$r, \@a, \$l); } push @a, $pm, $blib;Thread Next