[Scilab-Dev] building scialb 5.3.0 on OpenSolaris

Jonathan Blanchard BlanchardJ at ieee.org
Thu Jan 13 15:33:06 CET 2011


On Wed, Jan 12, 2011 at 5:02 AM, Sylvestre Ledru
<sylvestre.ledru at scilab.org> wrote:
> Le lundi 10 janvier 2011 à 14:08 -0400, Jonathan Blanchard a écrit :
>> On Mon, Jan 10, 2011 at 2:06 PM, Jonathan Blanchard <BlanchardJ at ieee.org> wrote:
>> > On Mon, Jan 10, 2011 at 1:37 PM, Jonathan Blanchard <BlanchardJ at ieee.org> wrote:
>> >> On Mon, Jan 10, 2011 at 1:18 PM, Kostas Oikonomou <ko at research.att.com> wrote:
>> >>> Jonathan Blanchard <BlanchardJ at ...> writes:
>> >>>
>> >>>>
>> >>>>
>> >>>> > Thanks.  I added '-lsunmath' to configure, and it gets added to the last
>> >>>> > compilation command, but the error still looks to be the same.
>> >>>> >
>> >>>> >
>> >>>>
>> >>>> This is a bit puzzling, are you using sun linker?
>> >>>>
>> >>>> Jonathan Blanchard
>> >>>
>> >>> Yes, /usr/ccs/bin/ld.
>> >>>
>> >>> I also searched on Google, and some of the undefined symbols, e.g.
>> >>> _gfortran_os_error are in the library /usr/lib/libgfortran.so. It seems to me
>> >>> that gfortran should include this library by default, but even after adding
>> >>> -lgfortran to LIBS, these symbols are still undefined.
>> >>>
>> >>> Kostas
>> >>>
>> >>>
>> >>
>> >> I'm trying a rebuild right now to compare build output. That -nostdlib
>> >> option in your link line is suspicious, I don't remember scilab
>> >> manually linking everything. In the meantime do you have CFLAGS,
>> >> LDFLAGS and friends defined to anything?
>> >>
>> >>
>> >> Jonathan Blanchard
>> >>
>> >
>> > k. Nevermind what I just said. The major diff between my link line and
>> > your is the linking of sunperf for that lib. Try applying the attached
>> > patch and then running autoreconf in your top source dir.
>> >
>> > Then slightly modify your build options by removing -lsunperf and
>> > -lsunmath from LIBS.
> Hello Jonathan,
>
> Do you want me to apply this patch ?
>
> Thanks
> S
>
> --
> -------------------------
> Sylvestre Ledru
> -------------------------
> The Scilab Consortium
> Digiteo
> Domaine de Voluceau
> Rocquencourt - B.P. 105
> 78153 Le Chesnay Cedex
> France
>
>

No, not really. I don't think building with sun performance library
using gcc is a really standard operation. Either way it's always
located deep into the Sun Studio install directory which also cannot
be easily guessed during configuration. IMO things should stay as they
are now.

Jonathan Blanchard



More information about the dev mailing list