Re: [Discuss-gnuradio] Make tests failed when Building for version 3.7.1git

[Tom Rondeau]

On Fri, Aug 23, 2013 at 10:32 AM, Marcus Leech <address@hidden> wrote:
> Is this a 32-bit install on 64-bit hardware issue?

Possibly. I was wondering about that, too. Thought we had fixed those
issues, but this might be a new version of it.

-- 
Tom
Visit us at GRCon13 Oct. 1 - 4
http://www.trondeau.com/grcon13

>
> on Aug 23, 2013, Tom Rondeau <address@hidden> wrote:
>
> On Fri, Aug 23, 2013 at 1:53 AM, Damon <address@hidden> wrote:
>>
>>> On Tue, Aug 20, 2013 at 9:05 AM, Damon <address@hidden> wrote:
>>>>>
>>>>> On Mon, Aug 19, 2013 at 8:29 AM, Damon <address@hidden>
>>>>> wrote:
>>>>>>
>>>>>> Hi all,
>>>>>>
>>>>>> I tried installing GNUradio v3.7.1git. The build completed, but
>>>>>> several
>>>>>> of
>>>>>> the build tests fail.
>>>>>>
>>>>>> The OS is Ubuntu 12.04 32bit, with kernel version Linux
>>>>>> 3.2.0-51-generic-pae. CPU is Intel® Core™ i7-3770 CPU @ 3.40GHz × 8 .
>>>>>>
>>>>>> The prerequisites were built by :
>>>>>> apt-get install libfontconfig1-dev libxrender-dev libpulse-dev swig
>>>>>> g++
>>>>>> \
>>>>>> automake autoconf libtool python-dev libfftw3-dev \
>>>>>> libcppunit-dev libboost1.48-all-dev libusb-dev libusb-1.0-0-dev fort77
>>>>>> \
>>>>>> libsdl1.2-dev python-wxgtk2.8 git-core \
>>>>>> libqt4-dev python-numpy ccache python-opengl libgsl0-dev \
>>>>>> python-cheetah python-lxml doxygen qt4-dev-tools libusb-1.0-0-dev \
>>>>>> libqwt5-qt4-dev libqwtplot3d-qt4-dev pyqt4-dev-tools python-qwt5-qt4 \
>>>>>> cmake git-core wget libxi-dev python-docutils gtk2-engines-pixbuf
>>>>>> r-base-dev
>>>>>> python-tk \
>>>>>> liborc-0.4-0 liborc-0.4-dev libasound2-dev python-gtk2
>>>>>>
>>>>>> then I ran cmake, make and make test.
>>>>>> The following tests FAILED:
>>>>>> 1 - qa_volk_test_all (Failed)
>>>>>> 85 - qa_freq_xlating_fir_filter (Failed)
>>>>>> 93 - qa_fir_filter (Failed)
>>>>>> 167 - qa_codec2_vocoder (Failed)
>>>>>> So I ran:
>>>>>> ctest --output-on-failure -O ctest_log_file
>>>>>> Attached please find the log file.
>>>>>>
>>>>>> How to fix it?
>>>>>>
>>>>>> Best regards,
>>>>>> Damon
>>>>>
>>>>> Damon,
>>>>>
>>>>> Thanks for the detailed report. Can you try uninstalling ORC and
>>>>> rebuilding without it? It looks like those errors might be related to
>>>>> differences in precision/rounding with ORC.
>>>>>
>>>>> And ignore the codec2 problem, unless you specifically want to use
>>>>> codec2. It's a known problem, but we don't have a fix for it, yet.
>>>>>
>>>> Hi Tom,
>>>>
>>>> There are 2 orc installed: liborc-0.4-0 and liborc-0.4-dev.
>>>> I uninstalled liborc-0.4-dev, rebuilt gnuradio. The volk passed
>>>> test. But several tests failed:
>>>>
>>>> 85 - qa_freq_xlating_fir_filter (Failed)
>>>> 93 - qa_fir_filter (Failed)
>>>> 167 - qa_codec2_vocoder (Failed)
>>>> Attached please find the log file: ctest_log_file.
>>>>
>>>> In my another computer, I uninstalled both liborc-0.4-0 and
>>>> liborc-0.4-dev.
>>>> Then I rebuilt uhd and gnuradio v3.7.0. Three tests mentioned above
>>>> failed
>>>> too.
>>>> Attached please find the log file: ctest_log_file2.
>>>>
>>>> Any suggestions?
>>>>
>>>> Thanks,
>>>> Damon
>>>>
>>> So again, let's not worry about the codec2 failure; that's something
>>> different. The two filter failures are probably related to the same
>>> issue since one uses the other.
>>>
>>> Can you try running volk_profile to make sure you Volk config file is
>>> created correctly? Then rerun those filter tests.
>>>
>> Hi Tom,
>>
>> After I have run volk_profile, more tests failed:
>>
>> 76 - test_gr_filter (Failed)
>> 83 - qa_fft_filter (Failed)
>> 85 - qa_freq_xlating_fir_filter (Failed)
>> 89 - qa_adaptive_fir_filter (Failed)
>> 93 - qa_fir_filter (Failed)
>> 125 - qa_cma_equalizer (Failed)
>> 132 - qa_ofdm_sync_sc_cfb (Failed)
>> 135 - qa_lms_equalizer (Failed)
>> 150 - qa_ofdm_txrx (Failed)
>> 167 - qa_codec2_vocoder (Failed)
>>
>> Attached please find the volk_config file and ctest_log_file.
>>
>> Regards,
>> Damon
>
>
> Strange as it sounds, those results actually look better. I'm not sure
> about the divide by zero case for the fft_filters, but the other
> results you have are very close to what they should be. It almost
> looks like there is an off-by-1 indexing problem. Either that or your
> ALU is messed up. But there's something strange about the numbers
> you're showing. And you haven't changed any files in the source code
> (that is, 'git status' shows nothing)?
>
> --
> Tom
> Visit us at GRCon13 Oct. 1 - 4
> http://www.trondeau.com/grcon13