bob@rancor.UUCP (Bob Willcox) (04/22/91)
I have been attempting to compile X386 1.1a on my ESIX 5.3.2D system.
Having been plaqued mostly with configuration problems at first I am now
to the point where I appear to be suffering from badly built shared
libs. The error message I get when make tries to link with any of the X
shared libraries built earlier is:
ld fatal: cannot combine NOLOAD input sections with other input sections;
possibly caused by shared library address space overlap
I am building this with gcc 1.39 with coff patches applied that I
obtained from suphys.
I am also getting errors compiling the at386 drivers with MANY messages
looking like this:
driver.c:68:base/index register must be 32 bit register
I have not yet tracked this one down, but appears that it may be coming
from the gcc-as assembler rather than the compiler.
Thanks for any help.
--
Bob Willcox ...!{rutgers|ames}!cs.utexas.edu!romp!rancor!bob
Phone: 512 258-4224roell@informatik.tu-muenchen.de (Thomas Roell) (04/22/91)
> > ld fatal: cannot combine NOLOAD input sections with other input sections; > possibly caused by shared library address space overlap > >I am building this with gcc 1.39 with coff patches applied that I >obtained from suphys. The bug is that GCC 1.39 as you configured uses GAS. Thus X386 won't compile properly. Use GCC 1.39 and the plain as. - Thomas -- _______________________________________________________________________________ E-Mail (domain): roell@lan.informatik.tu-muenchen.de UUCP (if above fails): roell@tumult.{uucp | informatik.tu-muenchen.de} famous last words: "diskspace - the final frontier..."
pim@cti-software.nl (Pim Zandbergen) (04/25/91)
roell@informatik.tu-muenchen.de (Thomas Roell) writes: >>I am building this with gcc 1.39 with coff patches applied that I >>obtained from suphys. >The bug is that GCC 1.39 as you configured uses GAS. Thus X386 won't compile >properly. Use GCC 1.39 and the plain as. Some time ago, a i386 binary version of a coff-patched gcc/gas combo was announced in comp.archives. It is claimed this version would be able to compile X386. Can anyone acknowledge this before I have it ftp'ed to me? Here's a part of the article. >Komplettes i386 gnu Compiler und Debugger Paket kann per anonymous ftp >abgeholt werden von >ftp.regent.e-technik.tu-muenchen.de >Files: >pub/gnu/gnu-i386/README >pub/gnu/gnu-i386/gnu-18.4.91.Z >Hier das README: >This is a complete binary package of gnu tools for the i386. >The tools are based on the dbx-in-coff patches by Michael Bloom. >With dbx-in-coff you gain the following: >- no need for library conversions and duplicate libraries. >- support of shared libraries. >- no more problems with forward references in debugging output. >- excellent g++ debugging including name demangling. >The package should run as is under Interactive and SCO Unix. [...] >If you compile X386 with this package, use the -mnobss switch to force >static uninitialized data into the data segment (I did not try to >compile X386 though, but I'm interested in your results). -- Pim Zandbergen domain : pim@cti-software.nl CTI Software BV uucp : uunet!mcsun!hp4nl!ctisbv!pim Laan Copes van Cattenburch 70 phone : +31 70 3542302 2585 GD The Hague, The Netherlands fax : +31 70 3512837