Re: [SOLVED][FORGE] struct package segmentation fault

[Juan Pablo Carbajal]

On Sun, Jan 26, 2014 at 8:33 PM, Juan Pablo Carbajal
<address@hidden> wrote:
> On Sun, Jan 26, 2014 at 7:50 PM, Olaf Till <address@hidden> wrote:
>> On Sun, Jan 26, 2014 at 07:11:14PM +0100, Juan Pablo Carbajal wrote:
>>> On Sun, Jan 26, 2014 at 5:37 PM, Olaf Till <address@hidden> wrote:
>>> > On Sun, Jan 26, 2014 at 03:51:40PM +0100, Juan Pablo Carbajal wrote:
>>> >>
>>> >> Olaf, Thank you for your time.
>>> >> This are the last lines of the configure output. I am compiling with no 
>>> >> flags.
>>> >>
>>> >>  Build Octave GUI:                   yes
>>> >>   JIT compiler for loops:             no
>>> >>   Build Java interface:               yes
>>> >>   Do internal array bounds checking:  no
>>> >>   Use octave_allocator:               no
>>> >>   Build static libraries:             no
>>> >>   Build shared libraries:             yes
>>> >>   Dynamic Linking:                    yes (dlopen)
>>> >>   Include support for GNU readline:   yes
>>> >>   64-bit array dims and indexing:     no
>>> >>
>>> >> I was executing Octave without the GUI. I can install the control
>>> >> package without crash, s yes, I believe is only with the struct
>>> >> package. Other packages install well also, geometry, io, general,
>>> >> signal, etc...
>>> >>
>>> >> There are two files cell2fields.oct
>>> >> /home/juanpi/.octave/struct-1.0.10/x86_64-pc-linux-gnu-api-v48+/cell2fields.oct
>>> >> (35K, 21. January)
>>> >> /home/juanpi/.octave/struct-1.0.10/x86_64-unknown-linux-gnu-api-v49+/cell2fields.oct
>>> >> (39K, 26 January (today))
>>> >
>>> > Nothing looks suspicious in the above. Though I didn't see a reason
>>> > anymore why this configuration should matter I've compiled Octave with
>>> > the same configuration now and could not reproduce the segfault. I
>>> > have no further ideas at the moment. As I said, it looks as if the
>>> > segfault happens when a function of Octave is called during loading of
>>> > the packages oct-files which actually should only be called for static
>>> > initialization at start of Octave.
>>> >
>>> > Olaf
>>> >
>>> > --
>>> > public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net
>>>
>>> Olaf do you know other pacage doing this. I tried the ones that I know
>>> they have compiled files and they work. What is special about
>>> struct...compared with control or geometry?
>>
>> I don't see anything special about the struct package. And no code of
>> the struct package should be run during installation, since there are
>> no PKG_ADD directives.
>>
>> You should be able to compile e.g. cell2fields.cc separately with
>> mkoctfile and load the resulting cell2fields.oct from Octaves current
>> working directory with 'help cell2fields'. But this could only give a
>> clue if the crash should not happen then.
>>
>> Olaf
>>
>> --
>> public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net
>
> The function compiles (with warnings) and then it is loadable and it works ok.
> So, is this something in Octave?

Ok, I found what is causing the crash. There was a version of
struct-1.0.10 already installed in the system.
Removing the old package fixed the issue. This might be a weakness in pkg.