Craig A. Berry wrote:
I am not sure where else to look for to try to isolate this.I've gone through essentially the same steps you have and not gotten any farther. I've attached a smaller reproducer test; it's still not really that small, but remove any line from it and you don't get the failure. There appears to be something about deleting a directory named 'blib' that corrupts the class object of the same name. We can band-aid over it by doing the following, which we perhaps should do at this point: --- lib/Module/Build/Platform/VMS.pm;-0 Thu Oct 25 04:50:40 2007 +++ lib/Module/Build/Platform/VMS.pm Wed Nov 14 07:42:00 2007 @@ -36,6 +36,7 @@ sub _set_defaults { $self->SUPER::_set_defaults(@_); $self->{properties}{build_script} = 'Build.com'; + $self->{properties}{blib} = '[.blib]'; }
I have to reluctantly agree with this. But I suspect that this is indicating that we have a buffer overrun condition somewhere that is probably VMS specific.
I took another glance at the rmdir code in VMS.C, but still do not see anything.
The test_type.t and xs.t tests are still failing for me when run under the test harness because they have verbose output that is not preceded by "#" characters.
I am going to see what I can find about ppm.t, to see if I can find out how the tarball is being created.
-John [EMAIL PROTECTED] Personal Opinion Only.
