RE: [Discuss-gnuradio] Fedora Core 6 on PS3

[Tom Rondeau]

> -----Original Message-----
> From: Eric Blossom [mailto:address@hidden
 
> On Mon, Mar 05, 2007 at 11:59:39PM -0500, Tom Rondeau wrote:
> > Alright, I've got my PS3 in my lab all set up with Fedora Core 6 (I
> > know, Bob, you told me to go with YDL, but I needed to play around
> > first). The installation went swimmingly until I tried to run a GNU
> > Radio app.
> >
> > I first tried some audio components, but it was having a hard time with
> > the audio card. I'm ignoring that and just want to prove I can get a
> > basic flow graph running. So I tried usrp_siggen.py, which lead to a
> > segmentation fault.
> >
> > A little digging showed that it occurs in basic_flow_graph.py in the
> > method _check_type_match when creating the connections. It seems it
> > doesn't like:
> > src_size = src_sig.sizeof_stream_item (src_endpoint.port)
> >
> > Some more digging showed that the d_sizeof_stream_item vector gets the
> > correct information pushed into it, but when it's read back out, it
> > looks like the value is just random data in memory.
> >
> > On the other hand, I ran a test using the new hier_block2's, and it
> > works just fine.
> >
> > So a few of questions:
> > 1. Anyone run Fedora Core 6 successfully?
> > 2. Should I try to use Fedora Core 5? (I have the ISO downloaded and
> > waiting)
> > 3. Is there a fix for this issue that someone can help me figure out?
> > 4. Should I just stick with hier_block2 graphs and force myself to
> > rewrite a number of the example codes with them?
> >
> > Thanks,
> > Tom
> 
> Hi Tom,
> 
> I've currently got FC5, FC6 and YDL installed on a 320GB 7200 RPM SATA
> drive.  (I'll post info on the setup on the wiki at some point.)

I've put instructions on my own internal website and will move it to the GNU
Radio Wiki (hopefully later today).

> I have yet to troubleshoot the Python QA failures.  Under FC5, they
> segfault.  From reports from Bob, I believe the situation is the same
> under YDL.  I haven't yet installed all the tool on FC6 and YDL.  The
> standalone test code works fine.  I'm not sure what's wrong, and
> haven't spent much time looking at the problem.

That's good to know. I just wanted confirmation of this stuff happening in
different builds and that it wasn't just me. It sounds like there are
general issues with Python in these systems, and from what I'm seeing, it's
almost like the memory is getting overwritten. 
 
> I wouldn't give up on FC6 yet, since I know the examples don't work on
> FC5.  Feel free to figure out what's going on ;)
> I'll find some later this week to look at it.

I'll keep looking, but I've got a number of other things to do this week and
just needed to prove that I can run something. If I can continue to develop
with the hier_block2 code, that's still a huge win in my opinion.

On that subject, I have example code to use the USRP to generate a sine wave
(a copy of usrp_siggen.py with hier_block2) that is currently running on my
PS3 :)

> I assume you built SWIG from source.  Have you tried "make check" on it?

At first, I just installed it with Yum, but that was my first instinct as to
where the problem may be. I installed it from source and just recompiled it
taking out all of the other language packages. Doing "make check" passed
most of the tests but failed on some java language issues. I haven't looked
any closer than this.

Tom