Re: [Discuss-gnuradio] constellation soft decoder

[Marcus Müller]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Phil, hi Miklos,

as far as I can tell, the default constructor of constellation doesn't
initialize d_pre_diff_code, which is correct (well, kind of correct)
considering that it does not set any constellation points, either.
So I guess both members are initialized with their default
constructor, i.e. empty std::vectors<int> or <complex>, respectively.

I don't really know how to elegantly fix this. I don't know if one
should remove the default constructor (and modify the depending
constellation_Xpsk constructors), or if that will break even more
things. Maybe the way to go would be having a method that sets the
constellation and automatically sets the pre_diff_code automatically
with the index range, just to prevent that error. This might however
be wasteful, for larger codes. Maybe one would just add a delegator
that either returns d_pre_diff_code[i] or just i, depending on the
existence of the predifferential codebook.

Anyway, I think Miklos is right: There is an error in calc_soft_dec in
using the pre_diff_code even in the case that it is not set or its
application disabled.

Miklos, would you be as nice as to file a bug for that? You're the one
who successfully diagnosed that. (You may, btw, simply think of
pre_diff_code simply as an index set of the constellation points)

Greetings,
Marcus

On 16.12.2013 14:29, Philip Balister wrote:
> On 12/16/2013 08:16 AM, Miklos Maroti wrote:
>> Hi Marcus,
>> 
>> Ok, that makes sense. It does not crash with QPSK, but it does
>> with BPSK and 8PSK, and BPSK and 8PSK do not set the
>> pre_diff_code. This explains why there is crash in one case and
>> not the other. Although neither sets the apply_pre_diff_code, so
>> even QPSK should not set pre_diff_code.
> 
> I took a quick look in Coverity for issues in gr-digital. It flags
> a bunch of uninitialized variables. I'll try and take a look at
> these this week.
> 
> There is an intermittent QA failure in some QA code and this might
> be an explanation.
> 
> Philip
> 
>> 
>> I think the real problem is in constellation::calc_soft_dec. I
>> does not check whether d_apply_pre_diff_code is set or not.
>> 
>> (By the way, I do not understand the usefulness of pre_diff_code.
>> Can you just permute the constellation points and get the same
>> effect?)
>> 
>> Miklos
>> 
>> On Mon, Dec 16, 2013 at 1:34 PM, Marcus Müller
>> <address@hidden> wrote: Well, after having a debug build and
>> then adding a lot of "volatile"s into the code at the crashing
>> function, it looks as if the d_pre_diff_code of the constellation
>> has size 0; and thus, you can't access the 0. (read: first)
>> element. How this could happen with a QPSK object is currently
>> subject to my confusion. Stay tuned for more confusing facts as
>> the code unfolds :).
>> 
>> Greetings, Marcus
>> 
>> On 16.12.2013 08:30, Miklos Maroti wrote:
>>>>> Hi Marcus,
>>>>> 
>>>>> I do not think it is a compiler bug, but something
>>>>> uninitialized. Constellation_qpsk and
>>>>> constellation_calcdist does not crash (I did not verify if
>>>>> they work). However, if I generate the LUT for 
>>>>> constellation_qpsk, then it crashes.
>>>>> 
>>>>> # uname -a Linux maroti-t410 3.11.0-14-generic #21-Ubuntu
>>>>> SMP Tue Nov 12 17:04:55 UTC 2013 x86_64 x86_64 x86_64
>>>>> GNU/Linux
>>>>> 
>>>>> # gcc --version gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1
>>>>> 
>>>>> Using latest master branch of gnuradio, hash
>>>>> 117058bac41c91, both CMAKE_BUILD_TYPE=Debug and Release
>>>>> produces the same behavior
>>>>> 
>>>>> # gdb `which python` core Reading symbols from 
>>>>> /usr/bin/python2.7...(no debugging symbols found)...done.
>>>>> 
>>>>> warning: core file may not match specified executable file.
>>>>> [New LWP 2249] [New LWP 2251] [New LWP 2240] [New LWP 2248]
>>>>> [New LWP 2247] [New LWP 2239] [New LWP 2250]
>>>>> 
>>>>> warning: Can't read pathname for load map: Input/output
>>>>> error. [Thread debugging using libthread_db enabled] Using
>>>>> host libthread_db library
>>>>> "/lib/x86_64-linux-gnu/libthread_db.so.1". Core was
>>>>> generated by `python ./constellation_simulation.py'. 
>>>>> Program terminated with signal 11, Segmentation fault. #0 
>>>>> 0x00007f98dbbe01a7 in
>>>>> gr::digital::constellation::calc_soft_dec ( this=0x48321c0,
>>>>> sample=..., npwr=<optimized out>) at 
>>>>> /home/mmaroti/workspace/gnuradio/gr-digital/lib/constellation.cc:286
>>>>>
>>>>>
>>
>>>>> 
286        for(int j = 0; j < k; j++) {
>>>>> 
>>>>> i stack #0  0x00007f98dbbe01a7 in 
>>>>> gr::digital::constellation::calc_soft_dec (
>>>>> this=0x48321c0, sample=..., npwr=<optimized out>) at 
>>>>> /home/mmaroti/workspace/gnuradio/gr-digital/lib/constellation.cc:286
>>>>>
>>>>>
>>
>>>>> 
#1  0x00007f98dbbe0f7e in gr::digital::constellation::soft_decision_maker (
>>>>> this=0x48321c0, sample=...) at 
>>>>> /home/mmaroti/workspace/gnuradio/gr-digital/lib/constellation.cc:355
>>>>>
>>>>>
>>
>>>>> 
#2  0x00007f98dbbe94dc in
>>>>> gr::digital::constellation_soft_decoder_cf_impl::work 
>>>>> (this=0x4fb11a0, noutput_items=8190, input_items=..., 
>>>>> output_items=...) at 
>>>>> /home/mmaroti/workspace/gnuradio/gr-digital/lib/constellation_soft_decoder_cf_impl.cc:68
>>>>>
>>>>>
>>
>>>>> 
#3  0x00007f98e1c5dd38 in gr::sync_interpolator::general_work (
>>>>> this=0x4fb11c0, noutput_items=<optimized out>,
>>>>> ninput_items=..., input_items=..., output_items=...) at 
>>>>> /home/mmaroti/workspace/gnuradio/gnuradio-runtime/lib/sync_interpolator.cc:67
>>>>>
>>>>>
>>>>> 
You should be able to reproduce the problem quite easily, but if
>>>>> needed I can provided a test program.
>>>>> 
>>>>> Best, Miklos
>>>>> 
>>>>> On Mon, Dec 16, 2013 at 7:59 AM, Marcus Müller
>>>>> <address@hidden> wrote:
>>>>>> Could you post the whole backtrace (as a github gist,
>>>>>> pastebin etc)? That would be very helpful, along with
>>>>>> information such as: Compiler (Gcc (version), VC++, clang
>>>>>> ...), Platform (Linux, win, mac,...), architecture
>>>>>> (x86(64), ARMv9, IBM S7390,...) and GNU Radio source
>>>>>> version.
>>>>>> 
>>>>>> Thanks and happy hacking, Marcus
>>>>>> 
>>>>>> On 12/16/2013 01:05 AM, Miklos Maroti wrote:
>>>>>>> 
>>>>>>> Hi Guys,
>>>>>>> 
>>>>>>> Constellation soft decoder crashes for
>>>>>>> constellation_bpsk() and constellation_8psk(). I could
>>>>>>> not find the cause of this problem, the back trace on
>>>>>>> my machine points to
>>>>>>> 
>>>>>>> gnuradio/gr-digital/lib/constellation.cc:286 for(int j
>>>>>>> = 0; j < k; j++) {
>>>>>>> 
>>>>>>> which does not make any sense. Any idea what could go
>>>>>>> wrong?
>>>>>>> 
>>>>>>> Miklos
>>>>>>> 
>>>>>>> _______________________________________________ 
>>>>>>> Discuss-gnuradio mailing list address@hidden 
>>>>>>> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> 
_______________________________________________ Discuss-gnuradio
>>>>>> mailing list address@hidden 
>>>>>> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>> 
>> _______________________________________________ Discuss-gnuradio
>> mailing list address@hidden 
>> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>> 
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.15 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSrwVRAAoJEAFxB7BbsDrLhd8H/0q7o3KX9iCShBn/X7VehxF2
LdR1peDL2VSqbGTmKnm6CltUs1yP7+bCAp+s6/TqTun7ticxDBSnh/0Drsc1jO2d
5E0UUKjjvuhdqKXfxbC/MGGSXlJ1Q1yGkkhmFvPS9ztQTT/Q3iLD8tFNshXdXlKc
FmPpyDWFDDD741AnqM4R5cXPnwUJkPV7aIvevYyrfYbjpSiK0t4yED0LVOZywnjs
aP2YB49qmWLpLYFAaIytcJF8cFkfumziPNQwpgndaZl8tuvwvUm7COHxXmBTdaqI
2IrB8gns+hR05uU9UAJdwaVI2gKGRB1hmP0MaHUxPd80ODfscTd5TAMxasBDWmE=
=HkF/
-----END PGP SIGNATURE-----