With the patches that I have submitted, the following tests are still
failing for me:
[-.ext.Cwd.t]cwd.t
I have not merged Craig's latest changes with mine as I am working on
getting the realpath() functionality in VMS, which is what this feature
really requires.
[-.lib.CPANPLUS.T]20_Cpanplus-Dist-MM.t
This is spawning a script to run Make Maker, with the instructions to
use PERL_ROOT[lib] as the include path. As PERL_ROOT is pointing to an
incompatible and older version of the include library, the test fails.
[-.lib.Module.Build.t]ppm.t
The TARBALL created or used by this test is being reported as corrupt,
both by this test, and by VMSTAR and gnutar. Reason for this corruption
is unknown at this time, but is blocking further testing of it.
[-.lib.Module.Build.t]test_type.t.
I suspect that the VMS pipe code is breaking up the output into multiple
lines, and that the test harness is expecting it as one long line. The
test passes when run interactively, but none of the very verbose
diagnostics are preceded by C<#>.
[-.lib.Module.Build.t]tilde.t
This now passes for me with the submitted patches, however when I
started the testing the logical name "HOME" was unexpected present with
a bad value in it. This caused the test to fail until I removed the
logical name.
TODO: [.VMS]TEST.COM needs to check, report, for any tests that leave
the following after they run, and repair it so that they do not break
subsequent tests. [This list probably will need expanding]
SYS$LOGIN:, HOME, PATH, or any files named ".;".
[-.lib.Module.Build.t]xs.t
The lookup of blib is failing unless this test is run in the Perl
debugger. I am not sure what is being clobbered. I have to find an
alternate way to debug this.
[-.lib.Test.Harness.t]failure.t
If I am interpreting the output of this test right when it is run
interactively, it is succeeding, so this needs more investigation as to
why it is failing when run under t/TEST. I suspect that the VMS pipe
code is adding an unexpected new line.
Of these, when I can start looking at this stuff again, my priorities
would be the realpath() stuff as it is needed for complete symbolic link
functionality, and then the xs.t and ppm.t failures. The xs.t and ppm.t
failures are indicating that something is bad in the VMS port of perl.
While I would prefer to concentrate on the new realpath() stuff, I
really do not like the two unknown failure cases in Module::Build.
-John
[EMAIL PROTECTED]
Personal Opinion Only
