svn commit: r319380 - head/math/libtommath

[b.f.]

On 5/30/13, Eitan Adler <eadler at freebsd.org> wrote:
> On 30 May 2013 19:47, b.f. <bf1783 at googlemail.com> wrote:
>> On 5/30/13, Eitan Adler <eadler at freebsd.org> wrote:
>>> On 30 May 2013 03:42, Martin Wilke <miwi at freebsd.org> wrote:
>>>>
>>>> On May 30, 2013, at 5:17 AM, Eitan Adler <eadler at FreeBSD.ORG> wrote:
>>>>
>>>>> On 29 May 2013 17:41, Martin Wilke <miwi at freebsd.org> wrote:
>>>>>> Author: miwi
>>>>>> Date: Wed May 29 15:41:00 2013
>>>>>> New Revision: 319380
>>>>>> URL: http://svnweb.freebsd.org/changeset/ports/319380
>>>>>>
>>>>>> Log:
>>>>>>  - Disable regression test
>>>>>>
>>>>>>  Reported by:  pointyhat building 2(D)-02:56:27
>>>>>
>>>>> This doesn't really explain what happened here.
>>>>> Did the regression test fail? Did the tests fail to build? Are there
>>>>> no regression tests?
>>>>>
>>>>> If the tests failed are they false positives, or do they indicate
>>>>> actual problems (where the port should be marked BROKEN)?
>>>>>
>>>>
>>>> the regression test hangs without timeout thats why i have disable it.
>>>> port it self builds fine no reason to mark it broken.
>>>
>>> Do the hangs indicate failure of the test program or failure of the
>>> actual program?
>>> If the latter it should be marked BROKEN.
>>
>> We don't automatically mark ports as BROKEN if they fail to pass a
>> test -- if we did, a great many more ports would be unavailable,
>> because most ports don't perform perfectly.  Often test failures are
>> caused by the conditions in which the tests are run, or are due to
>> problems in the testsuite, or represent uncommon or minor problems
>> that should ultimately be fixed, but don't render the port useless in
>> the meantime.
>
> Please read what I asked carefully.
>
> Is the failure an indication of a broken program?
>
> BROKEN is not used just because a port fails its regression tests but
> it isn't reserved just for build failures.

Of course it isn't. My point was that even if the test failure
indicated a "failure of the actual program", it doesn't necessarily
mean that the port should be marked BROKEN, as you seemed to suggest.
It depends on the nature of the failure.

b.