Re: [Discuss-gnuradio] problems building

[Tom Rondeau]

On Sat, Nov 9, 2013 at 8:50 AM, Tom Rondeau <address@hidden> wrote:
> On Sat, Nov 9, 2013 at 8:28 AM, Tom Rondeau <address@hidden> wrote:
>> On Sat, Nov 9, 2013 at 3:13 AM, Jeff Scaparra <address@hidden> wrote:
>>> I am trying to install gnuradio on Arch but I am having issues with
>>> qtgui and pfb_clock_sync. I use qt so I need to get this figured out.
>>>
>>> Any ideas?
>>>
>>> The following tests FAILED:
>>>         130 - qa_pfb_clock_sync (Failed)
>>>         166 - qa_qtgui (Failed)
>>>
>>> 151 130/179 Testing: qa_pfb_clock_sync
>>> 3152 130/179 Test: qa_pfb_clock_sync
>>> 3153 Command: "/usr/bin/sh"
>>> "/home/scap/AUR/gnuradio-git/src/gnuradio/build/gr-digital/python/digital/qa_pfb_clock_sync_test.sh"
>>> 3154 Directory:
>>> /home/scap/AUR/gnuradio-git/src/gnuradio/build/gr-digital/python/digital
>>> 3155 "qa_pfb_clock_sync" start time: Nov 09 03:08 EST
>>> 3156 Output:
>>> 3157 ----------------------------------------------------------
>>> 3158 .F
>>> 3159 ======================================================================
>>> 3160 FAIL: test02 (__main__.test_pfb_clock_sync)
>>> 3161 ----------------------------------------------------------------------
>>> 3162 Traceback (most recent call last):
>>> 3163   File 
>>> "/home/scap/AUR/gnuradio-git/src/gnuradio/gr-digital/python/digital/qa_pfb_clock_sync.py",
>>> line 139, in test02
>>> 3164     self.assertComplexTuplesAlmostEqual(expected_result, dst_data, 1)
>>> 3165   File 
>>> "/home/scap/AUR/gnuradio-git/src/gnuradio/gnuradio-runtime/python/gnuradio/gr_unittest.py",
>>> line 74, in assertComplexTuplesAl     mostEqual
>>> 3166     self.assertComplexAlmostEqual (a[i], b[i], places, msg)
>>> 3167   File 
>>> "/home/scap/AUR/gnuradio-git/src/gnuradio/gnuradio-runtime/python/gnuradio/gr_unittest.py",
>>> line 47, in assertComplexAlmostEq     ual
>>> 3168     (msg or '%s != %s within %s places' % (`first`, `second`, `places` 
>>> ))
>>> 3169 AssertionError: 1 != -1.0000078678131104 within 1 places
>>> 3170
>>> 3171 ----------------------------------------------------------------------
>>> 3172 Ran 2 tests in 0.029s
>>> 3173
>>> 3174 FAILED (failures=1)
>>> 3175 Using Volk machine: sse4_2_64_orc
>>> 3176 <end of output>
>>> 3177 Test time =   0.64 sec
>>> 3178 ----------------------------------------------------------
>>> 3179 Test Failed.
>>> 3180 "qa_pfb_clock_sync" end time: Nov 09 03:08 EST
>>> 3181 "qa_pfb_clock_sync" time elapsed: 00:00:00
>>>
>>> and
>>>
>>> 3895
>>> 3896 166/179 Testing: qa_qtgui
>>> 3897 166/179 Test: qa_qtgui
>>> 3898 Command: "/usr/bin/sh"
>>> "/home/scap/AUR/gnuradio-git/src/gnuradio/build/gr-qtgui/python/qtgui/qa_qtgui_test.sh"
>>> 3899 Directory:
>>> /home/scap/AUR/gnuradio-git/src/gnuradio/build/gr-qtgui/python/qtgui
>>> 3900 "qa_qtgui" start time: Nov 09 03:08 EST
>>> 3901 Output:
>>> 3902 ----------------------------------------------------------
>>> 3903 
>>> /home/scap/AUR/gnuradio-git/src/gnuradio/build/gr-qtgui/python/qtgui/qa_qtgui_test.sh:
>>> line 7: 25505 Segmentation fault      (core d     umped)
>>> /usr/bin/python2 -B
>>> /home/scap/AUR/gnuradio-git/src/gnuradio/gr-qtgui/python/qtgui/qa_qtgui.py
>>> 3904 <end of output>
>>> 3905 Test time =   0.42 sec
>>> 3906 ----------------------------------------------------------
>>> 3907 Test Failed.
>>> 3908 "qa_qtgui" end time: Nov 09 03:08 EST
>>> 3909 "qa_qtgui" time elapsed: 00:00:00
>>
>>
>> There were quite a few recent changes to both. But when posting
>> problems like this, please provide:
>>
>> - The version and architecture of the OS (32- or 64-bit)
>> - The version or git commit of GNU Radio you are working off of
>> - With qtgui issues, the version of Qt and Qwt you are using
>>
>> I'm trying this myself on various Linux VMs I have, both 64 and 32
>> bit, and not coming up with similar issues.
>>
>> Tom
>
>
> Found a system that has the problems you described. The pfb_clock_sync
> seems to be intermittent. It's locking at two different points and so
> is off by 1. That's not a huge problem, though it shouldn't be
> happening.
>
> The qtgui issue looks like it's with the qtgui.sink_c and qtgui.sink_f
> blocks. Everything else works fine. Working on a fix now.
>
> Tom


Ok, fixed it. The qtgui problem was really only a problem if we didn't
define the qApplication outside of a block, which is almost never the
case, so most applications would have been fine. But they changed a
few rules on me on how we have to treat that object. It's explained
more in the git commit and in comments in the constructors of the
qtgui sinks.

The pfb_clock_sync issue was mostly due to being too loose in one of
the parameters. I fixed the damping factor so we are now critically
damped (previously, it was seriously over-damped) and so convergence
is faster. I think there might have also been a small flaw in the
set_history value for the _fff version of the block (which is where
the failures were occurring).

Things should be behaving properly for you now. Thanks for the bug report.

Tom