[Ecls-list] ecl on Solaris/sparc?

Juan Jose Garcia-Ripoll juanjose.garciaripoll at googlemail.com
Wed Aug 27 08:01:42 UTC 2008


On Tue, Aug 26, 2008 at 4:28 AM, Raymond Toy <toy.raymond at gmail.com> wrote:
> First, the rest of ecl wants to build a 32-bit app, but gmp builds a
> 64-bit library, so that doesn't work together.  I think ecl may want to
> set ABI=32 when building gmp so that gmp will build a 32-bit library.  I
> didn't try making a 64-bit version of ecl, so I don't know if that would
> have worked or not.

First, I apologize for the lack of continued support of the solaris
platform, but since the Sourceforge compiler farm was closed, I have
had no luck finding another solaris account -- and installing
opensolaris on a virtual machine is, let's say, not really easy (help
is welcome).

Then, the mismatch between GMP's ABI and the operating system default
ABI is a know issue. That is why there is a flag --enable-slow-config
in ECL's configuration mechanism: it runs GMP's configuration before
even attempting to do ECL's, and then it uses the C and linker flags
from the resulting library.

> I stuffed the macros in ecl.h, and placed the C implementation in
> numbers.d.  I can provide a patch if desired.

Hmm, it should probably be much easier to force the compilation flag
that Michael suggested. There is a platform-dependent section in
aclocal.m4 which we can enlarge to include it.

> The last problem I have no solution for.  When linking in libeclgc, I
> get linker errors about non-aligned access in sparc_mac_dep.S.  Not sure
> what that means, but I'm basically stuck there.

That is the Boehm-Weiser garbage collector. Googling around shows that
this is due to Solaris shipping old versions of the assembler

http://osdir.com/ml/programming.garbage-collection.boehmgc/2005-07/msg00022.html

Juanjo

-- 
Instituto de Física Fundamental
CSIC, Serrano, 113, Madrid 28040 (Spain)
http://juanjose.garciaripoll.googlepages.com


More information about the ecl-devel mailing list