Re: GNU Octave 3.6.3 Released

[John W. Eaton]

On  5-Sep-2012, Philip Nienhuis wrote:

| 
| Jordi Gutiérrez Hermoso-2 wrote
| > 
| > On 5 September 2012 15:52, Sergei Steshenko <sergstesh@> wrote:
| > 
| > <snipped>
| > 
| > If that's your source directory and you built in-source there should
| > be a "run-octave" script in it along with test/fntests.m. If so,
| > please do
| > 
| >     ./run-octave -g --norc --silent --no-history test/fntests.m
| > 
| > This will put you in a gdb session. Do "run" to run the test suite.
| > This will run slower than normal because it'll be running in a
| > debugger. When Octave crashes, you should be dumped back to gdb. Do
| > "bt" to get the stack trace.
| > 
| > This isn't exactly one-button, but it's about as simple as it gets.
| > 
| 
| Continuing on my earlier post, along those (debug) lines I get (on Mageia 2)
| 
| octave:1> which pcg.m
| `pcg.m' is the file
| /home/philip/devel/oct363/../octave-3.6.3/scripts/sparse/pcg.m
| octave:2> test pcg.m
| 
| Program received signal SIGILL, Illegal instruction.
| 0xb4576acd in ATL_dJIK36x36x36TN36x36x0_a1_b0 () from
| /usr/lib/atlas-sse3/libatlas.so.3
| (gdb) bt
| #0  0xb4576acd in ATL_dJIK36x36x36TN36x36x0_a1_b0 () from
| /usr/lib/atlas-sse3/libatlas.so.3
| #1  0x0000007b in ?? ()
| #2  0x083474a0 in ?? ()
| Backtrace stopped: previous frame inner to this frame (corrupt stack?)
| (gdb) 
| 
| ... so I guess it is indeed Mageia that is at fault. 
| (And I also see why: my Pentium-M laptop's CPU has no SSE3. Apparently
| Mageia's CPU detection at install time has a hick-up).
| 
| Thanks Jordi, that was a very helpful hint!

I checked in the following changes which I think fix the segfault
problems in the Sparse.cc tests.

  http://hg.savannah.gnu.org/hgweb/octave/rev/f2e72944847b
  http://hg.savannah.gnu.org/hgweb/octave/rev/4663cc835c65

Could you please test these changes and confirm that they fix the
problem for you?

There will be a 3.6.4 release soon.

jwe