On Mon, 6 Aug 2001, Ralf Baechle wrote: > Date: Mon, 6 Aug 2001 11:45:56 +0200 > From: Ralf Baechle <ralf@oss.sgi.com> > To: John Heil <mipsdev@scsoftware.sc-software.com> > Cc: cobalt-22@devel.alal.com, cobalt-developers@list.cobalt.com, > linux-mips@oss.sgi.com > Subject: Re: Qube2 gcc 2.7.2 compiler error (fwd) > > On Sun, Aug 05, 2001 at 12:59:23PM +0000, John Heil wrote: > > Date: Sun, 5 Aug 2001 12:59:23 +0000 ( ) > > From: John Heil <mipsdev@scsoftware.sc-software.com> > > To: cobalt-22@devel.alal.com, cobalt-developers@list.cobalt.com, > > linux-mips@oss.sgi.com > > Subject: Qube2 gcc 2.7.2 compiler error (fwd) > > > > > > On the Qube2, gcc 2.7.2, option -s, to generate MIPS assembler > > corresponding to the input C code, generates invalid assembler > > by virtue of generating duplicate labels. The resultant > > assembler will not assemble, of course, due to the duplicate > > labels. The code (linux kernel's printk.c) compiles cleanly > > from C to object code. > > > > Q: How do I get valid assembler from gcc on Qube2 ? > > (My ultimate goal here is to be able to get listings out > > of gas.) > > gcc 2.7.2 creates a duplicate label for each function label. That is no > problem as both always have the same value. But I assume you're talking This is exactly the problem. The fact that the values are the same is causing the assembler interface to GCC to fail. When gcc -S outputs a .s assembler file, the GNU as assembler errors out on 'duplicate label'. Further, using the gcc -Wa,... without using -S, causes the compile to fail when the assembler is invoked and gives the same errors. > about a different type of label? Can you send a piece of small piece of > source code and the assembler code generated from it to demonstrate the > problem? > > Anyway, these days gcc 2.7.2 is so old these days it's not even funny. You > really should upgrade. > > Ralf > I'm happy to upgrade... What is the recommended level for kernel compiles and where can I find it. I am required to work on 2.0.34C53_SK using Qube2 for my build platform so I need compatibility. If I could cross compile on x86 that would be cool too. Thanx much John
