Christopher H. Laco wrote:
> chromatic wrote:
>> On Tuesday 02 September 2008 11:01:44 David Golden wrote:
>>
>>>> You encourage what you measure,
>>> In theory, yes. In practice, that hasn't been the experience to date.
>>> Testers over 70K:
>>>
>>> 1 587018 Chris Williams (BINGOS)
>>> 2 318527 Andreas J. König (ANDK)
>>> 3 188392 David Golden (DAGOLDEN)
>>> 4 151457 David Cantrell (DCANTRELL)
>>> 5 148505 Slaven Rezić (SREZIC)
>>> 6 73425 Jost Krieger (JOST)
>>> 7 73104 Yi Ma Mao (IMACAT)
>>> Do you think this group couldn't game the stats if all they wanted was
>>> a high score? Being snide about peoples volunteer efforts isn't
>>> particularly constructive.
>> Someone in that top seven has sent plenty of useless reports. ("Hi, I'm
>> from
>> CPAN Testers! I have my client configured not to install required
>> dependencies! Your distribution doesn't work! Hope that helps!")
>>
>>> If you think that people should be rewarded (acknowledged?) for
>>> "useful" reports, start defining "useful" and the heuristics you'd use
>>> to identify them.
>> * Does the report identify an actual failure for the common use case of CPAN
>> installation or does it identify a failure in configuring the CPAN Testers
>> client?
>>
>> * Does the report identify a known failure already reported elsewhere with
>> the
>> same characteristics?
>>
>> * Does the report identify a success on a previously unknown
>> platform/configuration combination?
>>
>> * Does the platform combination include a supported version of Perl?
>>
>> My criteria for usefulness suggest answers of "Yes. No. No. Yes. Yes."
>> I
>> realize that the third question is more difficult to answer in the presence
>> of XS components, but most of the distributions on the CPAN are pure Perl.
>>
>> -- c
>>
>>
>
>
> * Does the report actually include the error at all?
>
> useful: A CPAN testers FAIL report that actually includes the failure it
> signifies.
>
> http://www.nntp.perl.org/group/perl.cpan.testers/2008/08/msg2060496.html
> http://www.nntp.perl.org/group/perl.cpan.testers/2008/08/msg2060470.html
>
>
Yes, the second one does error in the middle of the output...barely.
Had the errors been after the 50k, the report would be doubly useless:
[Output truncated after 50K]
does no good when the sole purpose of such reports is to r1eport errors
from systems you aren't looking at, or even have locally to test with.
