>Why are we using kernel headers to build gdb? This header should come from
>glibc, and be the same for all Linux ports.
It's included by glibc's <sys/user.h>, which wants struct pt_regs. You won't
see the problem with kernel 2.2 because it doesn't define any PT_ constants in
the <asm/ptrace.h> file -- they're all in <linux/ptrace.h> and ptrace.c, the
former of which is already explicitly eschewed by user.h for exactly this
reason.
As well as fixing the kernel I think we should apply a patch like the one
below to glibc, to avoid any such problems in the future. (I haven't tested
this yet, but you get the idea.)
p.
2000-08-01 Philip Blundell <[EMAIL PROTECTED]>
* sysdeps/unix/sysv/linux/arm/sys/user.h: Avoid use of kernel
headers.
Index: sysdeps/unix/sysv/linux/arm/sys/user.h
===================================================================
RCS file: /cvs/glibc/libc/sysdeps/unix/sysv/linux/arm/sys/user.h,v
retrieving revision 1.2
diff -u -p -u -r1.2 user.h
--- user.h 1999/09/14 01:38:51 1.2
+++ user.h 2000/08/01 21:21:36
@@ -1,4 +1,4 @@
-/* Copyright (C) 1998, 1999 Free Software Foundation, Inc.
+/* Copyright (C) 1998, 1999, 2000 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -21,11 +21,6 @@
#include <features.h>
-/* <sys/ptrace.h> and <linux/ptrace.h> both define the PTRACE_* macros.
- This leads to compilation problems with programs which include both
- user.h and ptrace.h (eg: GDB). Do not include <linux/ptrace.h> here. */
-#include <asm/ptrace.h>
-
struct user_fp
{
struct fp_reg
@@ -44,9 +39,14 @@ struct user_fp
unsigned int init_flag;
};
+struct user_regs
+{
+ unsigned long int uregs[18];
+};
+
struct user
{
- struct pt_regs regs; /* General registers */
+ struct user_regs regs; /* General registers */
int u_fpvalid; /* True if math co-processor being used. */
unsigned long int u_tsize; /* Text segment size (pages). */
