[RFC] The never ending 32-bit vs. 64-bit blues

[RFC] The never ending 32-bit vs. 64-bit blues

[Mark Kettenis]

In the past we have had a lot of bug reports from people trying to
debug 64-bit programs with a native 32-bit gdb.  As far as I know we
don't support any native configuration that allows this.  I've never
quite managed to come up with a satisfactory solution to deal with
this.  Here's an attempt to deal with it by making configure complain.

Any opinions about this?

Mark


Index: configure.ac
===================================================================
RCS file: /cvs/src/src/gdb/configure.ac,v
retrieving revision 1.24
diff -u -p -r1.24 configure.ac
--- configure.ac	25 Jul 2005 15:08:40 -0000	1.24
+++ configure.ac	7 Nov 2005 22:34:19 -0000
@@ -34,6 +34,19 @@ AM_PROG_CC_STDC
 AC_CONFIG_AUX_DIR(`cd $srcdir;pwd`/..)
 AC_CANONICAL_SYSTEM
 
+# Sanity check; refuse to build a native GDB for a 64-bit target with
+# a 32-bit compiler.
+AC_CHECK_SIZEOF(void *)
+if test "${target}" = "${host}"; then
+  case "$host_cpu" in
+  hppa64|mips64*|powerpc64|sparc64|x86_64)
+    if test $ac_cv_sizeof_void_p != 8; then
+      AC_MSG_ERROR(
+[building native target ${host} with a 32-bit compiler is not supported])
+    fi
+  esac
+fi
+
 dnl List of object files and targets accumulated by configure.
 
 CONFIG_OBS=

Re: [RFC] The never ending 32-bit vs. 64-bit blues

[Jim Ingham]

Just as a point of information, Apple's PPC gdb (built 32 bit) can  
debug both 32 and 64 bit PPC apps.  We have a 32 bit native gdb with  
a 32 bit and a 64 bit gdbarch.  Seems to work okay so far.

Jim

On Nov 7, 2005, at 2:45 PM, Mark Kettenis wrote:

> In the past we have had a lot of bug reports from people trying to
> debug 64-bit programs with a native 32-bit gdb.  As far as I know we
> don't support any native configuration that allows this.  I've never
> quite managed to come up with a satisfactory solution to deal with
> this.  Here's an attempt to deal with it by making configure complain.
>
> Any opinions about this?
>
> Mark
>
>
> Index: configure.ac
> ===================================================================
> RCS file: /cvs/src/src/gdb/configure.ac,v
> retrieving revision 1.24
> diff -u -p -r1.24 configure.ac
> --- configure.ac    25 Jul 2005 15:08:40 -0000    1.24
> +++ configure.ac    7 Nov 2005 22:34:19 -0000
> @@ -34,6 +34,19 @@ AM_PROG_CC_STDC
>  AC_CONFIG_AUX_DIR(`cd $srcdir;pwd`/..)
>  AC_CANONICAL_SYSTEM
>
> +# Sanity check; refuse to build a native GDB for a 64-bit target with
> +# a 32-bit compiler.
> +AC_CHECK_SIZEOF(void *)
> +if test "${target}" = "${host}"; then
> +  case "$host_cpu" in
> +  hppa64|mips64*|powerpc64|sparc64|x86_64)
> +    if test $ac_cv_sizeof_void_p != 8; then
> +      AC_MSG_ERROR(
> +[building native target ${host} with a 32-bit compiler is not  
> supported])
> +    fi
> +  esac
> +fi
> +
>  dnl List of object files and targets accumulated by configure.
>
>  CONFIG_OBS=
>

Re: [RFC] The never ending 32-bit vs. 64-bit blues

[David Lecomber]

On Mon, 2005-11-07 at 23:45 +0100, Mark Kettenis wrote:
> In the past we have had a lot of bug reports from people trying to
> debug 64-bit programs with a native 32-bit gdb.  As far as I know we
> don't support any native configuration that allows this.  I've never
> quite managed to come up with a satisfactory solution to deal with
> this.  Here's an attempt to deal with it by making configure complain.
> 

Hi Mark,

AIX's can only be run within a 32-bit debugger, even for 64-bit apps --
ie. their ptrace only be called in 32-bit mode.

Cheers
d.

Re: [RFC] The never ending 32-bit vs. 64-bit blues

[Daniel Jacobowitz]

On Mon, Nov 07, 2005 at 11:45:49PM +0100, Mark Kettenis wrote:
> In the past we have had a lot of bug reports from people trying to
> debug 64-bit programs with a native 32-bit gdb.  As far as I know we
> don't support any native configuration that allows this.  I've never
> quite managed to come up with a satisfactory solution to deal with
> this.  Here's an attempt to deal with it by making configure complain.
> 
> Any opinions about this?

For the situation as you've stated it, I don't agree.  Not only did
Richard post some PPC Linux patches to make this work, I'll have
patches for it for MIPS Linux shortly; I've done all the kernel bits
already and most of the testing.

The patch as you've posted it, however, isn't really very useful for
that goal.  But it does seem useful for another.  When you build a
32-bit PowerPC debugger you should be configuring it for powerpc-linux;
if you configure it for powerpc64-linux but use -m32, you're likely to
get pretty surprising results.  So I have no problem with the patch
itself :-)

-- 
Daniel Jacobowitz
CodeSourcery, LLC