[rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[Ulrich Weigand]

Hello,

this is another small build fix that is a pre-requisite for the
next patch (which adds inf-ptrace.c on AIX).  The problem is that
on AIX, PTRACE_TYPE_ARG3 is a pointer, and trying to cast a
CORE_ADDR directly to it gives a compile warning.

The patch adds an intermediate cast to "long" to fix this problem;
there is precedent for this method in inf_ptrace_xfer_partial.

Tested on powerpc-aix-ibm5.3.0.0.

Bye,
Ulrich


ChangeLog:

	* inf-ptrace.c (inf_ptrace_fetch_register): Add intermediate cast
	to "long" before casting CORE_ADDR to PTRACE_TYPE_ARG3.
	(inf_ptrace_store_register): Likewise.


diff -urNp gdb-orig/gdb/inf-ptrace.c gdb-head/gdb/inf-ptrace.c
--- gdb-orig/gdb/inf-ptrace.c	Wed Apr 18 16:17:19 2007
+++ gdb-head/gdb/inf-ptrace.c	Mon Apr 23 21:14:20 2007
@@ -642,7 +642,7 @@ inf_ptrace_fetch_register (int regnum)
   for (i = 0; i < size / sizeof (PTRACE_TYPE_RET); i++)
     {
       errno = 0;
-      buf[i] = ptrace (PT_READ_U, pid, (PTRACE_TYPE_ARG3)addr, 0);
+      buf[i] = ptrace (PT_READ_U, pid, (PTRACE_TYPE_ARG3)(long)addr, 0);
       if (errno != 0)
 	error (_("Couldn't read register %s (#%d): %s."),
 	       REGISTER_NAME (regnum), regnum, safe_strerror (errno));
@@ -696,7 +696,7 @@ inf_ptrace_store_register (int regnum)
   for (i = 0; i < size / sizeof (PTRACE_TYPE_RET); i++)
     {
       errno = 0;
-      ptrace (PT_WRITE_U, pid, (PTRACE_TYPE_ARG3)addr, buf[i]);
+      ptrace (PT_WRITE_U, pid, (PTRACE_TYPE_ARG3)(long)addr, buf[i]);
       if (errno != 0)
 	error (_("Couldn't write register %s (#%d): %s."),
 	       REGISTER_NAME (regnum), regnum, safe_strerror (errno));
-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com

Re: [rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[Mark Kettenis]

> Date: Tue, 24 Apr 2007 23:37:24 +0200 (CEST)
> From: "Ulrich Weigand" <uweigand@de.ibm.com>
> 
> Hello,
> 
> this is another small build fix that is a pre-requisite for the
> next patch (which adds inf-ptrace.c on AIX).  The problem is that
> on AIX, PTRACE_TYPE_ARG3 is a pointer, and trying to cast a
> CORE_ADDR directly to it gives a compile warning.
> 
> The patch adds an intermediate cast to "long" to fix this problem;
> there is precedent for this method in inf_ptrace_xfer_partial.

I don't like this, but I guess this is unavoidable, I think it's more
correct to use uintptr_t instead of long though.

Mark

Re: [rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[Ulrich Weigand]

Mark Kettenis wrote:
> > Date: Tue, 24 Apr 2007 23:37:24 +0200 (CEST)
> > From: "Ulrich Weigand" <uweigand@de.ibm.com>
> > 
> > Hello,
> > 
> > this is another small build fix that is a pre-requisite for the
> > next patch (which adds inf-ptrace.c on AIX).  The problem is that
> > on AIX, PTRACE_TYPE_ARG3 is a pointer, and trying to cast a
> > CORE_ADDR directly to it gives a compile warning.
> > 
> > The patch adds an intermediate cast to "long" to fix this problem;
> > there is precedent for this method in inf_ptrace_xfer_partial.
> 
> I don't like this, but I guess this is unavoidable, I think it's more
> correct to use uintptr_t instead of long though.

I don't particularly like it either ...

Can we assume every system where inf-ptrace.c is built has uintptr_t?
I notice the type isn't frequently used elsewhere in GDB today.

If we decide to use uintptr_t, I guess the existing uses of long in
inf_ptrace_xfer_partial should be changed as well.


Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com

Re: [rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[Daniel Jacobowitz]

On Wed, Apr 25, 2007 at 02:26:26AM +0200, Ulrich Weigand wrote:
> Can we assume every system where inf-ptrace.c is built has uintptr_t?
> I notice the type isn't frequently used elsewhere in GDB today.

You Can Now.  Take a look at the generated gdb_stdint.h.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [rfc] [4/7] Modernize AIX target: inf-ptrace build fix

[Ulrich Weigand]

Daniel Jacobowitz wrote:
> On Wed, Apr 25, 2007 at 02:26:26AM +0200, Ulrich Weigand wrote:
> > Can we assume every system where inf-ptrace.c is built has uintptr_t?
> > I notice the type isn't frequently used elsewhere in GDB today.
> 
> You Can Now.  Take a look at the generated gdb_stdint.h.

Nice!  I'll use the following modified patch then.

Tested on powerpc-ibm-aix5.3.0.0.

Bye,
Ulrich


ChangeLog:

	* inf-ptrace.c: Include "gdb_stdint.h".
	(inf_ptrace_xfer_partial): Use "uintptr_t" instead of "long" as 
	intermediate type when casting CORE_ADDR to PTRACE_TYPE_ARG3.
	(inf_ptrace_fetch_register): Add intermediate cast to "uintptr_t"
	before casting CORE_ADDR to PTRACE_TYPE_ARG3.
	(inf_ptrace_store_register): Likewise.
	* Makefile.in (inf-ptrace.o): Update dependencies.

diff -urNp gdb-orig/gdb/inf-ptrace.c gdb-head/gdb/inf-ptrace.c
--- gdb-orig/gdb/inf-ptrace.c	2007-04-14 18:08:02.000000000 +0200
+++ gdb-head/gdb/inf-ptrace.c	2007-04-25 19:22:38.793800264 +0200
@@ -28,6 +28,7 @@
 #include "gdbcore.h"
 #include "regcache.h"
 
+#include "gdb_stdint.h"
 #include "gdb_assert.h"
 #include "gdb_string.h"
 #include "gdb_ptrace.h"
@@ -504,7 +505,8 @@ inf_ptrace_xfer_partial (struct target_o
 		    < rounded_offset + sizeof (PTRACE_TYPE_RET)))
 	      /* Need part of initial word -- fetch it.  */
 	      buffer.word = ptrace (PT_READ_I, pid,
-				    (PTRACE_TYPE_ARG3)(long)rounded_offset, 0);
+				    (PTRACE_TYPE_ARG3)(uintptr_t)
+				    rounded_offset, 0);
 
 	    /* Copy data to be written over corresponding part of
 	       buffer.  */
@@ -513,14 +515,16 @@ inf_ptrace_xfer_partial (struct target_o
 
 	    errno = 0;
 	    ptrace (PT_WRITE_D, pid,
-		    (PTRACE_TYPE_ARG3)(long)rounded_offset, buffer.word);
+		    (PTRACE_TYPE_ARG3)(uintptr_t)rounded_offset,
+		    buffer.word);
 	    if (errno)
 	      {
 		/* Using the appropriate one (I or D) is necessary for
 		   Gould NP1, at least.  */
 		errno = 0;
 		ptrace (PT_WRITE_I, pid,
-			(PTRACE_TYPE_ARG3)(long)rounded_offset, buffer.word);
+			(PTRACE_TYPE_ARG3)(uintptr_t)rounded_offset,
+			buffer.word);
 		if (errno)
 		  return 0;
 	      }
@@ -530,7 +534,8 @@ inf_ptrace_xfer_partial (struct target_o
 	  {
 	    errno = 0;
 	    buffer.word = ptrace (PT_READ_I, pid,
-				  (PTRACE_TYPE_ARG3)(long)rounded_offset, 0);
+				  (PTRACE_TYPE_ARG3)(uintptr_t)rounded_offset,
+				  0);
 	    if (errno)
 	      return 0;
 	    /* Copy appropriate bytes out of the buffer.  */
@@ -642,7 +647,7 @@ inf_ptrace_fetch_register (int regnum)
   for (i = 0; i < size / sizeof (PTRACE_TYPE_RET); i++)
     {
       errno = 0;
-      buf[i] = ptrace (PT_READ_U, pid, (PTRACE_TYPE_ARG3)addr, 0);
+      buf[i] = ptrace (PT_READ_U, pid, (PTRACE_TYPE_ARG3)(uintptr_t)addr, 0);
       if (errno != 0)
 	error (_("Couldn't read register %s (#%d): %s."),
 	       REGISTER_NAME (regnum), regnum, safe_strerror (errno));
@@ -696,7 +701,7 @@ inf_ptrace_store_register (int regnum)
   for (i = 0; i < size / sizeof (PTRACE_TYPE_RET); i++)
     {
       errno = 0;
-      ptrace (PT_WRITE_U, pid, (PTRACE_TYPE_ARG3)addr, buf[i]);
+      ptrace (PT_WRITE_U, pid, (PTRACE_TYPE_ARG3)(uintptr_t)addr, buf[i]);
       if (errno != 0)
 	error (_("Couldn't write register %s (#%d): %s."),
 	       REGISTER_NAME (regnum), regnum, safe_strerror (errno));
diff -urNp gdb-orig/gdb/Makefile.in gdb-head/gdb/Makefile.in
--- gdb-orig/gdb/Makefile.in	2007-04-25 19:30:42.240699372 +0200
+++ gdb-head/gdb/Makefile.in	2007-04-25 19:20:02.585131633 +0200
@@ -2170,7 +2170,7 @@ inflow.o: inflow.c $(defs_h) $(frame_h) 
 	$(serial_h) $(terminal_h) $(target_h) $(gdbthread_h) $(gdb_string_h) \
 	$(inflow_h) $(gdb_select_h)
 inf-ptrace.o: inf-ptrace.c $(defs_h) $(command_h) $(inferior_h) $(inflow_h) \
-	$(gdbcore_h) $(regcache_h) $(gdb_assert_h) \
+	$(gdbcore_h) $(regcache_h) $(gdb_stdint_h) $(gdb_assert_h) \
 	$(gdb_string_h) $(gdb_ptrace_h) $(gdb_wait_h) $(inf_child_h)
 infptrace.o: infptrace.c $(defs_h) $(command_h) $(frame_h) $(gdbcore_h) \
 	$(inferior_h) $(regcache_h) $(target_h) $(gdb_assert_h) \


-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com