At 10:48 PM -0700 6/2/03, Michael G Schwern wrote: >Since every test does: chdir($Is_VMS ? 'BFD_TEST_ROOT:[t]' : 't'); if >you'd change the BFD_TEST_ROOT definition in 00setup_dummy.t to already be >in t/ then we can squeeze out that last directory.
I'll take a look. I feel sure we've fixed this same issue more than once. > > And there are the wish-list items of having Install.pm and >> Manifest.pm be case blind when retrieving filenames from the file >> system but still being able to match those filenames to mixed-case >> package names. > >That's a big thing which I'm not going to worry about prior to release >unless a patch appears on my doorstep. If inspiration strikes I'll send something, but I think there are other priorities. > > On recent versions of VMS, all the ExtUtils tests are passing in >> bleadperl as long as we don't enable case preservation in filenames. > >They work when you *don't* enable it? That doesn't make sense. Sure it does. The C run-time traditionally (and still by default) downcases filenames, so, for example, "Foo.pm" becomes "foo.pm" when you learn about it via readdir(). Filenames acquired from manifests or package names have to be downcased to match, and this is what MM has always done on VMS. That gets me in trouble if I then flip a switch on the C run-time so that filename case is preserved; in this instance readdir() gives me "Foo.pm" and thus doesn't match the "foo.pm" I've stored in the expectation that names will be downcased. OK, you're right -- it doesn't make sense, but it's true :-). -- ________________________________________ Craig A. Berry mailto:[EMAIL PROTECTED] "... getting out of a sonnet is much more difficult than getting in." Brad Leithauser
