Re: [Boost-testing] [accumulators] spurious failures?

From: Martin Wille (mw8329_at_[hidden])
Date: 2008-01-09 08:15:14

John Maddock wrote:
> Richard Webb wrote:
>> Eric Niebler <eric <at> boost-consulting.com> writes:
>>
>>> Compile and link is fine. The run fails, but no reason is given. Any
>>> clues?
>>>
>> I just tried manually running the weighted_tail_mean.exe built by my
>> last test run, and it took over 5:30 to complete (though i dont know
>> why that alone would make the test fail, as i was running with a 20
>> minute timeout).
>>
>> On a connected note - i have a weighted_tail_mean.exe left over from
>> the accumulator tests, but no other exes (and the same is true of
>> regex_regress_threaded from the regex tests).
>> Is this intentional, or does it mean theres some other problem?

Yes, it is intentional.

> I've never fully understood why, but bjam deletes the .exe once the test is 
> complete: a real pain if you want to debug it afterwards!

Test executables get deleted only when the test succeeds.

Dave added that feature (at least partly due to my request) in order to
save space on the testing machines. The executable sizes sum up to more
than 15GB on my system. Deleting them saves a significant amount of
space, while regenerating them from the .o files can be done quickly.

Regards,
m