Solaris 10 5/08 Octave segfault

[Sergei Steshenko]

--- On Wed, 7/23/08, John Wiggins <jwiggins at rutgers.edu> wrote:

> From: John Wiggins <jwiggins at rutgers.edu>
> Subject: Re: Solaris 10 5/08 Octave segfault
> To: sergstesh at yahoo.com
> Cc: bug-octave at octave.org
> Date: Wednesday, July 23, 2008, 4:31 PM
> Hi Sergei,
> 
> Yes, I built gcc myself. I did the multistage bootstrap of
> the compiler 
> (including Ada), which more or less indicates the compiler
> is indeed 
> alive. Make check did pass a very large majority of the
> tests.
> 
> I also used the compiler to successfully build R, which
> suggests that 
> the fortran bits are working too.
> 
> thanks!
> -john
> 
> 
> Sergei Steshenko wrote:
> >
> > --- On Wed, 7/23/08, John Wiggins
> <jwiggins at rutgers.edu> wrote:
> >
> >   
> >> From: John Wiggins <jwiggins at rutgers.edu>
> >> Subject: Solaris 10 5/08 Octave segfault
> >> To: bug-octave at octave.org
> >> Date: Wednesday, July 23, 2008, 12:45 PM
> >> Hi,
> >>
> >> I just built octave-3.0.1 on Solaris 10 5/08
> sparc, but it
> >> segfaults on 
> >> startup:
> >>
> >> truss /usr/local/bin/octave-3.0.1
> >>
> >> ...
> >> .
> >> ..
> >> ..
> >> open64("/var/run/name_service_door",
> O_RDONLY)  =
> >> 4
> >> fcntl(4, F_SETFD, 0x00000001)                   =
> 0
> >> door_info(4, 0xFCBF1688)                        =
> 0
> >> door_call(4, 0xFFBFF060)                        =
> 0
> >> sysinfo(SI_HOSTNAME, "clean", 255)      
>        =
> >> 6
> >>     Incurred fault #6, FLTBOUNDS  %pc = 0xFCCB1A20
> >>       siginfo: SIGSEGV SEGV_MAPERR addr=0xFFFFFFF4
> >>     Received signal #11, SIGSEGV [default]I
> >>       siginfo: SIGSEGV SEGV_MAPERR addr=0xFFFFFFF4
> >>
> >>
> >> I built things with gcc/gfortran 4.3.1, 32-bit.
> >>
> >> Any ideas?
> >>
> >> -john
> >>
> >> _______________________________________________
> >> Bug-octave mailing list
> >> Bug-octave at octave.org
> >>
> https://www.cae.wisc.edu/mailman/listinfo/bug-octave
> >>     
> >
> > Did you build gcc-4.3.1 yourself ?
> >
> > If yes, did you run 'make check' for it ?
> There should be some failures,
> > but most of the tests should pass.
> >
> > Did you try to compile something simple with that gcc,
> like
> > "Hello, world !\n" just to make sure the
> compiler is "alive" ?
> >
> > Even though I've built gcc-4.3.1 (I am on IA32
> Linux), I am using gcc-4.2.4
> > for all may work, including octave and all its
> dependencies.
> >
> > FWIF, gcc-4.3.1 requires much stricter compliance with
> C/C++ standards.
> >
> > Regards,
> >   Sergei.
> >


Well, then I'm out of ideas.

Since I build octave using my AppsFronmScratch, I also build all its
dependencies except for e very thin system layer - things like X, etc.

Building dependencies I'm also running 'make check', so had the problem
like yours been due to a dependency, I would have caught it.

You may also try to build octave this way, though manually it should be
very painful.

I again suggest to build gcc-4.2.4 and try building octave using it.

It's easier than building octave + dependencies, and you'll know whether
gcc is the problem.

Regards,
  Sergei.