Then you've been bitten by Devel::Autoflush.

This has been a necessary evil (IMHO) to address a perl bug/feature
where print() without a newline to STDOUT flushes on a STDIN read when
STDOUT is a terminal, but doesn't when STDOUT is a pipe (as happens
under CPAN::Reporter).  There are a lot of Makefile.PL programs that
do that, unfortunately.

However, I'm now convinced that it breaks too many tests, either due
to $|=1 itself or due to PERL5OPT being set at all.  The next release
of CPAN::Reporter will limit Devel::Autoflush to only
Makefile.PL/Build.PL and will not use it during testing.

-- David

On Tue, Oct 7, 2008 at 12:22 AM, Chris Dolan <[EMAIL PROTECTED]> wrote:
> So, it turns out I was wrong.  It wasn't taint that was causing my test
> failures, but instead $|=1.  I added more logging and Slaven's smoke machine
> revealed that, actually.
> Chris
>
> On Oct 6, 2008, at 5:26 PM, Slaven Rezic wrote:
>
>> "David Golden" <[EMAIL PROTECTED]> writes:
>>
>>> One issue is that taint can be a per-test-file setting from the
>>> shebang line in a *.t file.
>>>
>>> That said, if a tester is running everything under taint, we should
>>> probably capture that.
>>
>> How one would do this --- using PERL5OPT=-t ?
>> Such a system would never generate a PASS report...
>>
>> Regards,
>>    Slaven
>>
>>>
>>> -- David
>>>
>>> On Mon, Oct 6, 2008 at 12:32 PM, Barbie <[EMAIL PROTECTED]> wrote:
>>>>
>>>> Hi Chris,
>>>>
>>>> I'm forwarding this to the [email protected] mailing list,
>>>> as it really covers Test-Reporter, CPAN-Reporter and CPANPLUS-YACSmoker.
>>>>
>>>> Cheers,
>>>> Barbie.
>>>>
>>>> On Sun, Oct 05, 2008 at 11:38:19PM -0500, Chris Dolan wrote:
>>>>>
>>>>> Hi Barbie,
>>>>>
>>>>> I noticed today that the cpantester reports don't indicate whether
>>>>> the tester's smoke environment runs under taint mode or not.  I just
>>>>> fixed a long-standing bug in one of my modules that only happened
>>>>> when the tests were run under -T (my mistake, I know).  It could help
>>>>> if the taint setting could be reported, perhaps near the $^X value.
>>>>>
>>>>> Please let me know if I should file this somewhere on RT or an email
>>>>> list instead of just to you.
>>>>>
>>>>> Thanks much.  I'm a big fan of the cpantester community and am
>>>>> grateful for your hard work.
>>>>> Chris
>>>>
>>>> --
>>>> Birmingham Perl Mongers <http://birmingham.pm.org>
>>>> Memoirs Of A Roadie <http://barbie.missbarbell.co.uk>
>>>>
>>>>
>>>>
>>>
>>
>> --
>> Slaven Rezic - slaven <at> rezic <dot> de
>>
>>    tktimex - time recording tool
>>    http://sourceforge.net/projects/ptktools/
>
>

Reply via email to