Technically, the issue (at least within CPAN.pm) is that each phase
has something like this:

    elsif ( $self->_should_report('make') ) {
        my ($output, $ret) = CPAN::Reporter::record_command($system);
        CPAN::Reporter::grade_make( $self, $system, $output, $ret );
        $system_ok = ! $ret;
    }

The output is captured, used for a grade and then thrown away.
Instead it would need to be persisted, perhaps only for XS modules,
until the test phase where it would be prepended to the test output in
the input to grade_test().  And it would need to be discarded at some
point to avoid ballooning memory usage.

David



On Tue, Jun 3, 2014 at 1:16 PM, David Oswald <daosw...@gmail.com> wrote:
> I could be making a mistake getting into this conversation, but here's one
> of my observations:
>
> There have been times when I have wished that I could obtain the build
> reports for an XS module.  It seems that the shortcoming is there is no way
> (to my knowledge) for the module author to cause the diagnostic output
> during "make" to be dumped anywhere aside from STDOUT/STDERR.  If one could
> specify an EU::MM, or EU::ParseXS option that caused 'make' to tee output to
> a file, a module author could write a test that reads the file and dumps it
> to "diag" or "note" messages, which will appear in reports.
>
>
> On Tue, Jun 3, 2014 at 6:08 AM, David Golden <x...@xdg.me> wrote:
>>
>> On Tue, Jun 3, 2014 at 4:17 AM, demerphq <demer...@gmail.com> wrote:
>> > On top of the point that this policy makes very little sense for an XS
>> > module, the fact is that the *build* step failed, yet cpantesters gave
>> > me
>> > *test* results. Sounds like the cpan tester framework is broken. Why did
>> > it
>> > run tests when the make step failed?
>>
>> Did the make step error?  I.e. did it return a non-zero exit code?  If
>> so, then it's a bug
>>
>> > And I have to say I really find that this policy undermines the utility
>> > of
>> > cpantesters for me.
>>
>> It's not "policy" its just how it's been done.  Remember we used to
>> spam perl.org with test reports?  And remember that someone pays for
>> the storage that CPAN Testers uses to store the millions of reports it
>> has.
>>
>> BTW, have you or Booking donated to CT lately?  It would be much
>> appreciated so there are funds to continue to reimburse me for the AWS
>> bill?
>>
>> http://www.enlightenedperl.org/donations.html
>>
>> I agree that for XS modules, keeping the build results in the report
>> would be very helpful.  Right now, there's no trivial way to do that.
>> If you'd like to open a CPAN-Reporter ticket, I don't know when I'll
>> get to it, but at least the idea won't be lost.
>>
>> > With cpantesters I have an automated build system that tells me my stuff
>> > is
>> > broken, but does not give me the details I need to fix it, and when I
>> > ask
>> > for them apparently it is too much bother for the person running the
>> > tests
>> > to actually supply what I asked.
>>
>> I've found most testers are pretty responsive to a request to provided
>> detailed output to help diagnose an unusual failure.
>>
>> --
>> David Golden <x...@xdg.me> Twitter/IRC: @xdg
>
>
>
>
> --
>
> David Oswald
> daosw...@gmail.com



-- 
David Golden <x...@xdg.me> Twitter/IRC: @xdg

Reply via email to