Re: segfault in tests

[Jaroslav Hajek]

On Fri, Feb 20, 2009 at 3:06 AM, Ben Abbott <address@hidden> wrote:
>
> On Feb 19, 2009, at 1:04 PM, John W. Eaton wrote:
>
>> On 19-Feb-2009, Jaroslav Hajek wrote:
>>
>> | On Thu, Feb 19, 2009 at 6:49 PM, John W. Eaton <address@hidden> wrote:
>> | > After updating today, I see this crash when running tests:
>> | >
>> | >  scripts/general/accumarray.m ...........................panic:
>> Segmentation fault -- stopping myself...
>> | > make[2]: *** [check] Segmentation fault
>> | > make[2]: Leaving directory `/scratch/jwe/build/octave/test'
>> | > make[1]: *** [check] Error 2
>> | > make[1]: Leaving directory `/scratch/jwe/build/octave'
>> | > make: *** [check] Error 2
>> | >
>> | > Does this happen for anyone else, or is it just me?
>> | >
>> | > jwe
>> | >
>> |
>> | Hmm,
>> | I now see it too. Strangely enough, running just "test sortrows" seems
>> | to run fine. I'll try to investigate what is wrong.
>>
>> I just did
>>
>>  ./run-octave -g
>>
>> and "test accumarray" and it crashes here:
>>
>>  (gdb) where
>>  #0  0x00007f3e1aeaeed5 in raise () from /lib/libc.so.6
>>  #1  0x00007f3e1aeb03f3 in abort () from /lib/libc.so.6
>>  #2  0x00007f3e1aeeb3a8 in ?? () from /lib/libc.so.6
>>  #3  0x00007f3e1aef0948 in ?? () from /lib/libc.so.6
>>  #4  0x00007f3e1aef2a56 in free () from /lib/libc.so.6
>>  #5  0x00007f3e21701422 in ~ArrayRep (this=0x11bc7d0) at
>> /home/jwe/src/octave/liboctave/Array.h:86
>>  #6  0x00007f3e217014fd in ~Array (this=0x7fff2a513be0) at
>> /home/jwe/src/octave/liboctave/Array.cc:84
>>  #7  0x00007f3e21466b7f in F__sort_rows_idx__ (address@hidden) at
>> /home/jwe/src/octave/src/data.cc:5672
>>  #8  0x00007f3e218d3a7a in octave_builtin::do_multi_index_op
>> (this=0x9253b8, nargout=1, address@hidden)
>>     at /home/jwe/src/octave/src/ov-builtin.cc:107
>>  #9  0x00007f3e218d3d56 in octave_builtin::subsref (this=0x9253b8,
>> address@hidden, address@hidden,
>>     nargout=1) at /home/jwe/src/octave/src/ov-builtin.cc:55
>>
>> But test sortrows succeeds.  I guess the sortrow tests don't include
>> the same kind of call that accumarray is doing.
>>
>> jwe
>
> I also encountered the seg-fault during "make check".
>
> Running from the command line, I first get a failure (no seg-fault). Running
> a second time produces the seg-fault.
>
> Perhaps there is some hint as to what is wrong in the failure?
>
> octave:1> test accumarray
>  ***** assert (accumarray
> ([1,1;2,1;2,3;2,1;2,3],101:105,[2,4],@max,NaN),[101,NaN,NaN,NaN;104,NaN,105,NaN])
> !!!!! test failed
> assert (accumarray ([1, 1; 2, 1; 2, 3; 2, 1; 2, 3], 101:105, [2, 4], @max,
> NaN),[101, NaN, NaN, NaN; 104, NaN, 105, NaN]) expected
>   101   NaN   NaN   NaN
>   104   NaN   105   NaN
> but got
>   101   NaN   NaN   NaN
>   104   NaN   NaN   NaN
> NaNs don't matchoctave:2> test accumarray
> panic: Segmentation fault -- stopping myself...
> /usr/local/bin/octave-3.1.51+: line 2: 56349 Segmentation fault
>  /Users/bpabbott/Development/mercurial/octave-3.1.51/run-octave
>
> Ben
>
>

As it happens,
in the morning the problem was suddenly crystally clear. Fixing things
up yesterday, I broke something else :(
I checked in a fix and slightly simplified and ooized the design.

http://hg.savannah.gnu.org/hgweb/octave/rev/89b95972e178

cheers

-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz