Someone *finally* followed up to my bug report that I filed against Fedora Core 1 regarding Test::Harness being part of the released distro but in actuality it fails the build tests, and should NOT have been released. I don't know how this gets past the rpm's rebuild unless it's NOT doing a make test, but anyway....
Someone at long last responded to the bug report. Bugzilla id 118221
However I don't know anything about the Perl header internals. I've advised Andy Lester of the small amount of progress with the report, however if there's anyone out there more familiar with the Perl header internals than I who would care to look at the follow up to the report and perhaps advise them whether this suggestion (comment #2) is safe to do, and perhaps help the Fedora Core people determine what's causing the problem to begin with, it would be much appreciated.
their comment:
The problem seems to arise from the perl header
/usr/lib/perl5/5.8.3/i386-linux-thread-multi/bits/typesizes.ph --
the definition of __FSID_TYPE translates badly into perl:
\'struct struct\' { \'int\' $__val[2]; }
This is not a valid perl expression. It's trying to duplicate
Fedora's definition of the type, struct { int __val[2]; }. I
commented out the offending line, and Test::Harness builds without
warning. I don't know if there's any harm in removing the
definition, nor do I know what it should look like.Re:Not a Test::Harness bug.
rafael on 2004-08-27T22:22:25
1. if you want to see how the rpm has been built, look at the spec file in the srpm (e.g. does it make a "make test" or not)
2. the perl 5.8.5 that comes in mandrakelinux 10.1 beta 2 has a patched h2ph (by me:) that hopefully avoids this error. Anyway the fedora perl packager will probably need to patch by hand the .ph files. The h2ph docs warn against this :)