[PATCH] i386bsd-nat.c tweak

[PATCH] i386bsd-nat.c tweak

[Mark Kettenis]

This removes the consistency checks for really old BSDs.  We don't
really support those and it hurts GNU/k*BSD.

Committed.


Index: ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>
 
	* i386bsd-nat.c (_initialize_i386bsd_nat): Remove `struct
	sigcontext' offset consistency checks for ancient BSD.

 
Index: i386bsd-nat.c
===================================================================
RCS file: /cvs/src/src/gdb/i386bsd-nat.c,v
retrieving revision 1.29
diff -u -p -r1.29 i386bsd-nat.c
--- i386bsd-nat.c 31 May 2004 11:20:41 -0000 1.29
+++ i386bsd-nat.c 9 Aug 2004 20:31:18 -0000
@@ -357,10 +357,10 @@ _initialize_i386bsd_nat (void)
 #define SC_REG_OFFSET i386nbsd_sc_reg_offset
 #elif defined (OpenBSD)
 #define SC_REG_OFFSET i386obsd_sc_reg_offset
-#else
-#define SC_REG_OFFSET i386bsd_sc_reg_offset
 #endif
 
+#ifdef SC_REG_OFFSET
+
   /* We only check the program counter, stack pointer and frame
      pointer since these members of `struct sigcontext' are essential
      for providing backtraces.  More checks could be added, but would
@@ -410,4 +410,6 @@ Please report this to <bug-gdb@gnu.org>.
     }
 
   SC_FP_OFFSET = offset;
+
+#endif /* SC_REG_OFFSET */
 }

Re: [PATCH] i386bsd-nat.c tweak

[Robert Millan]

On Mon, Aug 09, 2004 at 10:48:25PM +0200, Mark Kettenis wrote:
> diff -u -p -r1.29 i386bsd-nat.c
> --- i386bsd-nat.c 31 May 2004 11:20:41 -0000 1.29
> +++ i386bsd-nat.c 9 Aug 2004 20:31:18 -0000
> @@ -357,10 +357,10 @@ _initialize_i386bsd_nat (void)
>  #define SC_REG_OFFSET i386nbsd_sc_reg_offset
>  #elif defined (OpenBSD)
>  #define SC_REG_OFFSET i386obsd_sc_reg_offset
> -#else
> -#define SC_REG_OFFSET i386bsd_sc_reg_offset
>  #endif

This disables the code below on GNU/kFreeBSD, but it's still enabled in
FreeBSD 4.x or later.

Our sigcontext structure is like FreeBSD's (the code is borrowed from
machine/signal.h).  My impression is that on GNU/kFreeBSD we want to define
SC_REG_OFFSET i386fbsd4_sc_reg_offset.  Please let me know if I'm missing
something.

Build on GNU/kFreeBSD still works, but it might be that keeping SC_REG_OFFSET
undefined results in a weird runtime bug that will be difficult to trace, so
I'd really like to be sure this part is ok.
 
> +#ifdef SC_REG_OFFSET
> +
>    /* We only check the program counter, stack pointer and frame
>       pointer since these members of `struct sigcontext' are essential
>       for providing backtraces.  More checks could be added, but would
> @@ -410,4 +410,6 @@ Please report this to <bug-gdb@gnu.org>.
>      }
>  
>    SC_FP_OFFSET = offset;
> +
> +#endif /* SC_REG_OFFSET */
>  }

-- 
Robert Millan

(Debra and Ian) (Gnu's Not (UNiplexed Information and Computing System))/\
(kernel of *(Berkeley Software Distribution))

Re: [PATCH] i386bsd-nat.c tweak

[Mark Kettenis]

   Date: Tue, 10 Aug 2004 00:44:24 +0200
   From: Robert Millan <rmh@debian.org>

   On Mon, Aug 09, 2004 at 10:48:25PM +0200, Mark Kettenis wrote:
   > diff -u -p -r1.29 i386bsd-nat.c
   > --- i386bsd-nat.c 31 May 2004 11:20:41 -0000 1.29
   > +++ i386bsd-nat.c 9 Aug 2004 20:31:18 -0000
   > @@ -357,10 +357,10 @@ _initialize_i386bsd_nat (void)
   >  #define SC_REG_OFFSET i386nbsd_sc_reg_offset
   >  #elif defined (OpenBSD)
   >  #define SC_REG_OFFSET i386obsd_sc_reg_offset
   > -#else
   > -#define SC_REG_OFFSET i386bsd_sc_reg_offset
   >  #endif

   This disables the code below on GNU/kFreeBSD, but it's still enabled in
   FreeBSD 4.x or later.

Yes indeed.

   Our sigcontext structure is like FreeBSD's (the code is borrowed from
   machine/signal.h).  My impression is that on GNU/kFreeBSD we want to define
   SC_REG_OFFSET i386fbsd4_sc_reg_offset.  Please let me know if I'm missing
   something.

Your sigcontext structure better be identical to FreeBSD's since its
part of the kernel ABI.

   Build on GNU/kFreeBSD still works, but it might be that keeping
   SC_REG_OFFSET undefined results in a weird runtime bug that will be
   difficult to trace, so I'd really like to be sure this part is ok.

It will be caught on normal FreeBSD.

Mark

Re: [PATCH] i386bsd-nat.c tweak

[Robert Millan]

On Thu, Aug 12, 2004 at 10:55:23AM +0200, Mark Kettenis wrote:
> 
>    This disables the code below on GNU/kFreeBSD, but it's still enabled in
>    FreeBSD 4.x or later.
> 
> Yes indeed.
> 
> [...]
> 
>    Build on GNU/kFreeBSD still works, but it might be that keeping
>    SC_REG_OFFSET undefined results in a weird runtime bug that will be
>    difficult to trace, so I'd really like to be sure this part is ok.
> 
> It will be caught on normal FreeBSD.

We have this check:

  #if defined (__FreeBSD_version) && __FreeBSD_version >= 400011
  #define SC_REG_OFFSET i386fbsd4_sc_reg_offset

This ensures that SC_REG_OFFSET is defined for either 4.x or 5.x (i.e. any
non-obsolete version of FreeBSD).  If not defining it triggers a bug, how
can it be caught on normal FreeBSD, if that situation never happens?

Perhaps what you want is something like this?:

  #if defined (__FreeBSD_version) && __FreeBSD_version >= 400011 \
  	__FreeBSD_version < 500000
  #define SC_REG_OFFSET i386fbsd4_sc_reg_offset

Then i386fbsd4_sc_reg_offset would be for 4.x only, which seems like the
intended result.  Is that it?

-- 
Robert Millan

(Debra and Ian) (Gnu's Not (UNiplexed Information and Computing System))/\
(kernel of *(Berkeley Software Distribution))