This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
An egcs bug or linux bug?
- To: jinbo21 at soback dot kornet dot nm dot kr
- Subject: An egcs bug or linux bug?
- From: hjl at lucon dot org (H.J. Lu)
- Date: Thu, 21 May 1998 11:29:42 -0700 (PDT)
- Cc: egcs at cygnus dot com, torvalds at transmeta dot com (Linus Torvalds), linux-kernel at vger dot rutgers dot edu (linux kernel)
>
> Hellow,
>
> I met following error while compiling linux kernel-2.1.103 with
> egcs-19980517 patched with your egcs-980519.ppro.diff and
> egcs-980519.ll.diff.
>
> my environment: rh 5.0(glibc-2.0.93+latest patches), your egcs-1.0.3,
> binutils-2.9.1.0.4
> ASUS P/I-P55T2P4 rev. 3.1 (bios 0206.004, pipelined-
> burst SRAM 512K)
> intel pentium MMX 166(changed to 83x2)
> 64M samsung EDO RAM
>
> It might be another problem of egcs-19980517(alias?) or come from
> my overclocked CPU. But, anyhow, I report this to you.
>
> BTW, egcs-1.0.3 compiles kernel-2.1.103 well.
>
> .....
> make[2]: Leaving directory `/usr/src/linux-2.1.103/lib'
> make[1]: Leaving directory `/usr/src/linux-2.1.103/lib'
> make[1]: Entering directory `/usr/src/linux-2.1.103/arch/i386/kernel'
> ....
> gcc -D__KERNEL__ -I/usr/src/linux-2.1.103/include -Wall \
> -Wstrict-prototypes -O3 -fomit-frame-pointer -pipe -mcpu=pentium \
> -march=pentium -malign-loops=0 -malign-jumps=0 -malign-functions=0 \
> -Dpentium -c -o bios32.o bios32.c
> ../../egcs-19980517/gcc/alias.c:475: Internal compiler error \
> in function rtx_equal_for_memref_p
> cpp: output pipe has been closed
> make[1]: *** [bios32.o] Error 1
> make[1]: Leaving directory `/usr/src/linux-2.1.103/arch/i386/kernel'
> make: *** [linuxsubdirs] Error 2
>
> Thanks,
>
Here is a small test case.
# gcc -B/home/work/build/bin/egcs/gcc/ -S -O2 foo.c
/home/work/build/src/egcs/gcc/alias.c:474: Internal compiler error in function rtx_equal_for_memref_p
I am also enclosing a patch for Linux kernel 2.1.103. But I am not
if it is an egcs bug or linux kernel bug. Even if it is a linux bug,
egcs should give out a better message.
BTW, you should use -mcpu=pentiumpro -march=pentiumpro with my
PPro patches.
Thanks.
H.J.
----foo.c---
int check_pcibios(void)
{
unsigned int signature, eax, ebx, ecx;
unsigned long flags, pcibios_entry;
__asm__ ("lcall (%%edi)\n\t"
"jc 1f\n\t"
"xor %%ah, %%ah\n"
"1:"
: "=d" (signature),
"=a" (eax),
"=b" (ebx),
"=c" (ecx)
: "1" (0xb101 ),
"D" (&pcibios_entry)
: "memory");
__asm__ __volatile__("pushl %0 ; popfl": :"g" ( flags ):"memory") ;
if (signature)
return 1;
return 0;
}
Index: arch/i386/kernel/bios32.c
===================================================================
RCS file: /home/work/cvs/linux/linux/arch/i386/kernel/bios32.c,v
retrieving revision 1.1.1.18
diff -u -r1.1.1.18 bios32.c
--- bios32.c 1998/05/21 15:10:28 1.1.1.18
+++ bios32.c 1998/05/21 18:23:22
@@ -543,7 +543,7 @@
pci_indirect.address = pcibios_entry + PAGE_OFFSET;
save_flags(flags); cli();
- __asm__("lcall (%%edi)\n\t"
+ __asm__ __volatile__ ("lcall (%%edi)\n\t"
"jc 1f\n\t"
"xor %%ah, %%ah\n"
"1:"