t/04_client_tests.t fails with perl 5.21.7

[eserte]

See http://matrix.cpantesters.org/?dist=Riak-Light for a pass/fail overview and http://www.cpantesters.org/cpan/report/d64550d2-8a73-11e4-8b82-bb06e1bfc7aa for a sample fail report. I don't see a simple reason why it's failing. If you think this is a perl regression, then please report at perl5-porters.

[peczenyj]

Hello,

I am reading the code and seems the problem is related to the global
variable $! or some cast to string. this is "ugly" but in this case, make
sense

I will try to reproduce with perlbrew, then check if there is some
workaround to fix the test ( maybe my code must be more consistent, I saw
other failures like this :
http://www.cpantesters.org/cpan/report/33eda834-8a73-11e4-9774-8304e1bfc7aa
and seems i should use strerror in many places - to be careful with
locale/language settings ).

Then I will try to write a small proof of concept and sent to the
perl5-porters to see if it is a bug or not.

Thanks

Tiago

On Thu, Dec 25, 2014 at 6:28 PM, Slaven Rezić [email protected]
wrote:

See http://matrix.cpantesters.org/?dist=Riak-Light for a pass/fail
overview and
http://www.cpantesters.org/cpan/report/d64550d2-8a73-11e4-8b82-bb06e1bfc7aa
for a sample fail report. I don't see a simple reason why it's failing. If
you think this is a perl regression, then please report at perl5-porters.

—
Reply to this email directly or view it on GitHub
#18.

Tiago B. Peczenyj
Linux User #405772

http://about.me/peczenyj

[peczenyj]

Hello

I write this minimal test and check under 5.12.2 and 5.21.7

package Subject;
use Data::Dumper;

sub new {
  bless {};
}

sub ping {
  die "lol ($!)"
}

package main;

use strict;
use warnings;
use Data::Dumper;
use Test::More;
use Test::Exception;
use POSIX qw(ETIMEDOUT strerror);


        my $client = Subject->new();

        $! = ETIMEDOUT;
        #debug
        diag Dumper { ETIMEDOUT => ETIMEDOUT, error => $! };

        my $errmsg = strerror(ETIMEDOUT);
        throws_ok { $client->ping() } qr/lol \($errmsg\)/,
          "should die";

done_testing;

results:

$ perlbrew switch 5.12.2
$ prove -lv b.t
b.t .. 
ok 1 - should die
1..1
# $VAR1 = {
#           'error' => 'Operation timed out',
#           'ETIMEDOUT' => 60
#         };
ok
All tests successful.
Files=1, Tests=1,  0 wallclock secs ( 0.03 usr  0.00 sys +  0.04 cusr  0.01 csys =  0.08 CPU)
Result: PASS
$ perlbrew switch 5.21.7
$ prove -lv b.t
b.t .. 
not ok 1 - should die
1..1
# $VAR1 = {
#           'ETIMEDOUT' => 60,
#           'error' => 'Operation timed out'
#         };

#   Failed test 'should die'
#   at b.t line 29.
# expecting: Regexp ((?^:lol \(Operation timed out\)))
# found: lol () at b.t line 9.
# Looks like you failed 1 test of 1.
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/1 subtests 

Test Summary Report
-------------------
b.t (Wstat: 256 Tests: 1 Failed: 1)
  Failed test:  1
  Non-zero exit status: 1
Files=1, Tests=1,  0 wallclock secs ( 0.03 usr  0.01 sys +  0.08 cusr  0.01 csys =  0.13 CPU)
Result: FAIL

IF we move the debug message to the line below POSIX::strerror, we discover the $! variable has been erased.

And yet, the strerror has been changed between 5.21.6 and 5.21.7

Thanks for the report.

[eserte]

An even smaller test case:

$ perl5.21.6 -MErrno=ETIMEDOUT -MPOSIX=strerror -e '$! = ETIMEDOUT; strerror(ETIMEDOUT); warn $!'         
Connection timed out at -e line 1.
$ perl5.21.7 -MErrno=ETIMEDOUT -MPOSIX=strerror -e '$! = ETIMEDOUT; strerror(ETIMEDOUT); warn $!'
Warning: something's wrong at -e line 1.

And I see the problem in POSIX.pm: the strerror() implementation loads first locale.pm, which clobbers $!, and then does a local $!. Probably worth a bug report at perl5-porters.

A simple workaround would be to just "require locale" first.

[peczenyj]

I open one ticket

Thanks

https://rt.perl.org/Ticket/Display.html?id=123503
Em 26/12/2014 11:34, "Slaven Rezić" [email protected] escreveu:

An even smaller test case:

$ perl5.21.6 -MErrno=ETIMEDOUT -MPOSIX=strerror -e '$! = ETIMEDOUT; strerror(ETIMEDOUT); warn $!'
Connection timed out at -e line 1.
$ perl5.21.7 -MErrno=ETIMEDOUT -MPOSIX=strerror -e '$! = ETIMEDOUT; strerror(ETIMEDOUT); warn $!'
Warning: something's wrong at -e line 1.

And I see the problem in POSIX.pm: the strerror() implementation loads
first locale.pm, which clobbers $!, and then does a local $!. Probably
worth a bug report at perl5-porters.

A simple workaround would be to just "require locale" first.

—
Reply to this email directly or view it on GitHub
#18 (comment).

[peczenyj]

this error is related to a bugus implementation of POSIX::strerror on this particular version. I will close this ticket and follot the 123503 ticket on perl.org

[peczenyj]

Seems it was fixed https://rt.perl.org/Public/Bug/Display.html?id=123503