[root@rouzic]:/usr/src/Math-Pari-2.001804# PERL_DL_NONLAZY=1 /usr/bin/
+perl -Iblib/arch -Iblib/lib -I/usr/local/lib/perl5/5.6.1/i686-linux -
+I/usr/local/lib/perl5/5.6.1 -e 'use Test::Harness qw(&runtests $verbo
+se); $verbose=0; runtests @ARGV;' t/analyz.t
t/analyz............Variable in perl2pari is not of known type at test
+_eng/Testout.pm line 8.
        ...propagated at t/analyz.t line 5.
t/analyz............dubious
        Test returned status 255 (wstat 65280, 0xff00)
FAILED--1 test script could be run, alas--no output ever seen
[download]

So let's have a look at t/analyz.t and test_eng/Testout.pm and see what is going on

t/anaylz.t

  $file = __FILE__;
  $file =~ m|([^/.]*)\.t$| or die;
  @ARGV = "pari-2.1.1/src/test/32/$1";
  do 'test_eng/Testout.pm';
  die if $@;
[download]

Ok, this just appears to be doing an indirection to PARI's own regression tests.

test_eng/Testout.pm

     1  #! perl
     2  #       $rcs = ' $Id: testout.t,v 1.2 1997/09/22 10:13:37 ilya
+ Exp ilya $ ' ;
     3
     4  use Math::Pari qw(:DEFAULT pari_print :all);
     5  use vars qw($x $y $z $k $t $q $a $u $j $l $name $other $n);
     6  die "Need a path to a testout file" unless @ARGV;
     7
     8  $file = shift;
     9  {
    10    open TO, "< $file" or die "open `$file': $!";
    11    local $/ = "\n? ";
    12    @tests = <TO>;
    13    close TO or die "close: $!";
    14  }
    [...]
[download]

Here I drew a blank. I don't understand how the error message "Variable in perl2pari is not of known type at test_eng/Testout.pm line 8." relates to the above code. Line 8 is just saying $file = shift;

I don't know if you've ever tried to run the debugger on a regression test, but after a bit of spelunking I was able to see that all Test::Harness does is to say words to the effect of

  $cmd = '/usr/bin/perl -w t/analyz.t|'
  my $fh = new FileHandle;
  $fh->open($cmd) or print "can't run $test. $!\n";
[download]

That sounds easy enough, I think I'll try that myself. This time I've got it.

  Too few args 0 for PARI function shift at test_eng/Testout.
        ...propagated at t/analyz.t line 5.
[download]

Okay, so there's some sneaky overloading going on... what am I to do about it? Another thing that dampens my spirits it that the CPAN Testers rate the module as a FAIL on both ppc-linux and sparc-openbsd platforms. Certainly I have nothing so exotic, but right now I'd appreciate any pointers to help me solve my problem. I must say that the README and INSTALL files are particularly useless with respect to troubleshooting.

So, tilly, how did you do it?

g r i n d e r