imacat <[EMAIL PROTECTED]> writes:
> Dear Shlomi Fish,
>
> Hi. This is imacat from Taiwan. Sorry for the reply delay.
>
> On Tue, 04 Nov 2008 11:21:53 +0200
> Shlomi Fish <[EMAIL PROTECTED]> wrote:
>> The testing of http://nntp.x.perl.org/group/perl.cpan.testers/2539942 for
>> Error.pm fails. It seems like it exits on t/pod-coverage.t . One problem may
>> be the extremely long $PERL5LIB.
>
> The long PERL5LIB is resulted from a long dependency chain of another
> module. The dependency chain is too long to track the original testing
> module. I spent a couple of days but still cannot find the original
> testing module and reproduce this failure. I have submit 2 revised PASS
> reports.
>
I am not convinced that the PERL5LIB was too long in this case. The
test output in the report is cropped, unfortunately, but the
environment section below shows that PERL5LIB has just a size of about
20k, which should not cause a problem on modern Unix systems (on
Windows systems however, this would already hit some limits).
But the size of the failing open3 command is already >50k, and it is
even cropped, so it's probably much longer! Pick up one of the
directories in this command line, say GD-2.41-6xIk4L/blib/arch: it
appears three times. It looks like something is overflowing the
maximum command line on this system, which is probably 256k or so.
My guess is that Test::Harness is causing this problem. Test::Harness
needs to do some magic if the test file is flagged with -T, because
PERL5LIB is tainted, but does it wrong somehow, adding the libraries
too many times into the command line. The easiest workaround is to
remove the -T flag, but I think Test::Harness needs a fix here.
Regards,
Slaven
--
Slaven Rezic - slaven <at> rezic <dot> de
tkrevdiff - graphical display of diffs between revisions (RCS, CVS or SVN)
http://ptktools.sourceforge.net/#tkrevdiff
