Opening a ticket to track the multiple fails from CPAN Testers
for the PDL-2.4.10_003 release. All the fails for this test were
on amd64 platforms and OS freebsd version 9.0 and openbsd
version 5.0.
Of the reports with PASS (including t/gd_oo_tests.t) there
were some amd64 platforms but the archname *never*
included amd64. E.g.:
Bad platforms:
6 osname=freebsd, osvers=9.0-release, archname=amd64-freebsd
8 osname=freebsd, osvers=9.0-release, archname=amd64-freebsd-thread-multi
5 osname=openbsd, osvers=5.0, archname=OpenBSD.amd64-openbsd
8 osname=openbsd, osvers=5.0, archname=OpenBSD.amd64-openbsd-thread-multi
Good platforms (with either BSD or AMD64 in the osname= line):
1 osname=freebsd, osvers=8.2-release, archname=i386-freebsd
2 osname=freebsd, osvers=8.2-release, archname=i386-freebsd-64int
3 osname=freebsd, osvers=8.2-release, archname=i386-freebsd-thread-multi
3 osname=freebsd, osvers=8.2-release, archname=i386-freebsd-thread-multi-64int
5 osname=gnukfreebsd, osvers=8.1-1-amd64, archname=x86_64-gnukfreebsd
4 osname=gnukfreebsd, osvers=8.1-1-amd64, archname=x86_64-gnukfreebsd-thread-multi
2 osname=linux, osvers=2.6.18-4-xen-amd64, archname=x86_64-linux-thread-multi
1 osname=linux, osvers=3.2.0-2-amd64, archname=x86_64-linux-thread-multi
5 osname=openbsd, osvers=5.0, archname=OpenBSD.i386-openbsd
5 osname=openbsd, osvers=5.0, archname=OpenBSD.i386-openbsd-64int
3 osname=openbsd, osvers=5.0, archname=OpenBSD.i386-openbsd-thread-multi
4 osname=openbsd, osvers=5.0, archname=OpenBSD.i386-openbsd-thread-multi-64int
Re-enabled these tests for git and CPAN developer releases.
David Mertens reported a failure with the gd_oo tests on the PDL-porters list http://mailman.jach.hawaii.edu/pipermail//pdl-porters/2012-March/004538.html, I also see those same failures on my Mac (a segfault after 'ok 19').
Closing this ticket as the problem reports are no longer
being generated for PDL-2.4.11. This may have been a
platform/system bug and not one from PDL.
Marking the gd_oo_tests.t as TODO with the hopes of tracking down the specific problem.
For the upcoming PDL-2.4.12 release, the failing tests will be skipped if the AUTOMATED_TESTING variable is set. Otherwise, the tests will be run---and most likely fail for BSD OSes on AMD64 platforms. Presumably, the user in the non-automated case could determine whether or not a single test file failure of t/gd_oo_tests.t warrants override for install in spite of the fail.
Ideally, if they can reproduce the problem, I invite them to help debug the problem. Marking this as priority 1 since no actual users have reported the problem and the conditional skip will prevent massive PDL CPAN Tester failures for all BSD+AMD64 platforms.
This may be fixed with commit 7bbab7. Please check, and if it works we can un-TODO the gd_oo_tests and see if the BSD AMD64 testers can pass it now.
Problem appears to be resolved. Closing this ticket.