#432 t/gd_oo_tests.t fails with perl 5.25.12

[Slaven Rezic]

The test suite fails with perl 5.25.12 (various operating systems: freebsd 10, linux debian jessie & stretch, ubuntu 16.04, centos7; both unthreaded and threaded builds):

#   Failed test 'incorrectly initialize an object from an inline hash: non-string key'
# 
#   at t/gd_oo_tests.t line 269.
# 
# Looks like you failed 1 test of 32.
# 
t/gd_oo_tests.t ................. 
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/32 subtests 

I don't see this failure wth perl 5.25.11 or earlier, so it could be a perl regression.

[Andreas Koenig]

Could it be a GD/PDL regression? On the smoker without libgd I have no problem with PDL, on the other smoker with libgd I fail this test regardless which perl version.

[Slaven Rezic]

It seems to be a problem with the pure perl implementation of List::MoreUtils.

Works as expected:

$ perl5.18.4 -Mblib -MPDL::IO::GD -e 'warn PDL::IO::GD->new(filename=>"/home/eserte/src/bbbike/images/car.gif", [34]=>12)'
PDL::IO::GD::new couldn't parse its arguments.
Expected a hash-ref or an inline hash or just a filename

Is causing the test failure:

$ env LIST_MOREUTILS_PP=1 perl5.18.4 -Mblib -MPDL::IO::GD -e 'warn PDL::IO::GD->new(filename=>"/home/eserte/src/bbbike/images/car.gif", [34]=>12)'
PDL::IO::GD=HASH(0x801ca9468) at -e line 1.

For some reason the XS implementation of List::MoreUtils does not get build for perl 5.25.12, hence the impression that the test fails only for 5.25.12.

[Slaven Rezic]

Actually it seems to be a problem with the usage of @_ in the test script:

$ env LIST_MOREUTILS_PP=1 perl5.18.4 -MList::MoreUtils=none -e '@_ = ([34]=>12); warn none { ref $_[2*$_] } 0'     
1 at -e line 1.
$ env LIST_MOREUTILS_PP=1 perl5.18.4 -MList::MoreUtils=none -e '@x = ([34]=>12); warn none { ref $x[2*$_] } 0'
0 at -e line 1.

[Petr Písař]

That never worked with pure-Perl List::MoreUtils. The none is defined as:

sub none (&@)
{
    my $f = shift;
    foreach (@_)
    {
        return 0 if $f->();
    }
    return 1;
}

That means the @_ is changed into empty list when entering $f->() subroutine. It works in the XS implementation because the $$f is called directly on each argument list item.

I'm not expert on lexical scoping but @_ seems to be a special case in contrast to other named arrays.

[Slaven Rezic]

For some reason the XS implementation of List::MoreUtils does not get build for perl 5.25.12, hence the impression that the test fails only for 5.25.12.

I just realized that List::MoreUtils::XS is a separate distribution which may or may not be installed.

[Petr Písař]

A work-around to work with pure-Perl List::MoreUtils.

[Chris Marshall]

Thanks for the patch and catching the bitrot. I haven't been able to monitor the CPAN Testers status properly due to issues with the database and ones with the web display which, when it has data, often will fail to show any test results even when the summary says otherwise.

[Chris Marshall]

The patch has been applied and in the CPAN developers release CHM/PDL-2.017_01.tar.gz which will be released as PDL-2.018 once correct operation is confirmed.

[kmx]

I would like to see this released, any chance to have PDL-2.018?

[Chris Marshall]

It looks like CPAN Testers reports are coming in again and that the latest
developer release is good. @kmx: is a release this weekend soon enough for
you (there is one more thing I'm working to finish for 2.018)?

On Mon, May 15, 2017 at 4:14 AM, kmx kmx1@users.sf.net wrote:

I would like to see this released, any chance to have PDL-2.018?

• [bugs:#432] https://sourceforge.net/p/pdl/bugs/432/ t/gd_oo_tests.t
  fails with perl 5.25.12*

Status: pending-fixed
Group: normal
Created: Fri Apr 21, 2017 07:18 AM UTC by Slaven Rezic
Last Updated: Sun Apr 30, 2017 06:43 PM UTC
Owner: Chris Marshall

The test suite fails with perl 5.25.12 (various operating systems: freebsd
10, linux debian jessie & stretch, ubuntu 16.04, centos7; both unthreaded
and threaded builds):

Failed test 'incorrectly initialize an object from an inline hash: non-string key'

at t/gd_oo_tests.t line 269.

Looks like you failed 1 test of 32.

t/gd_oo_tests.t .................
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/32 subtests

I don't see this failure wth perl 5.25.11 or earlier, so it could be a
perl regression.

---

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/pdl/bugs/432/

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/

[kmx]

This weekend would be absolutely great, thanks.