Re: PATCH: Workaround IRIX 5 compilation problem

[Rainer Orth <ro@TechFak.Uni-Bielefeld.DE>]

Andrew Cagney writes:

> Does the attached also work?  This avoids blindly un/redefining _KMEMUSER.

Test in progress.

Two comments: you may want to keep the comment about the IRIX 5
<sys/proc.h> workaround in configure.in.

> Index: configure.in
> ===================================================================
> RCS file: /cvs/src/src/gdb/configure.in,v
> retrieving revision 1.151
> diff -u -r1.151 configure.in
> --- configure.in	25 Apr 2004 20:42:45 -0000	1.151
> +++ configure.in	28 Apr 2004 21:16:44 -0000
> @@ -632,7 +632,10 @@
>    *-*-solaris2.[[6789]])
>        AC_DEFINE(NEW_PROC_API)
>        ;;
> -  esac
> +  mips-sgi-irix5*)
> +      AC_DEFINE([_KMEMUSER], 1)
> +      ;;
> + esac

Lost one space before esac :-)

I've already found one other (completely unrelated) problem: this time, I'm
building with gcc 3.5 from CVS: the build failed here:

/vol/gnu/src/gdb/gdb-6.1/sim/mips/interp.c: In function `mips_core_signal':
/vol/gnu/src/gdb/gdb-6.1/sim/mips/interp.c:2326: error: invalid lvalue in assignment
/vol/gnu/src/gdb/gdb-6.1/sim/mips/interp.c:2334: error: invalid lvalue in assignment
make[2]: *** [interp.o] Error 1

This is due to http://gcc.gnu.org/gcc-3.4/changes.html (The cast-as-lvalue
extension has been removed for C++ and deprecated for C and Objective-C.)

#define COP0_BADVADDR ((unsigned32)(COP0_GPR[8]))

The assignments to COP0_BADVADDR are hard errors now.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University