[commit] New function gdbarch_infi_fill

[commit] New function gdbarch_infi_fill

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 215 bytes --]

Hello,

The attached adds the new function gdbarch_info_fill.  This new function 
fills in un-initialized fields of the "info" structure using global 
"set" variables and the "gdbarch" parameter.

committed,
Andrew

[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 7250 bytes --]

2003-11-09  Andrew Cagney  <cagney@redhat.com>

	* Makefile.in (arch-utils.o): Update dependencies.
	* arch-utils.c: Include "osabi.h".
	(gdbarch_info_fill): New function.
	* arch-utils.h (gdbarch_info_fill): Declare.
	* gdbarch.sh (gdbarch_update_p): Call "gdbarch_info_fill".
	* gdbarch.c: Re-generate.

Index: arch-utils.c
===================================================================
RCS file: /cvs/src/src/gdb/arch-utils.c,v
retrieving revision 1.102
diff -u -r1.102 arch-utils.c
--- arch-utils.c	9 Nov 2003 12:07:16 -0000	1.102
+++ arch-utils.c	9 Nov 2003 23:42:45 -0000
@@ -31,6 +31,8 @@
 #include "gdb_assert.h"
 #include "sim-regno.h"
 
+#include "osabi.h"
+
 #include "version.h"
 
 #include "floatformat.h"
@@ -678,6 +680,54 @@
   memset (info, 0, sizeof (struct gdbarch_info));
   info->byte_order = BFD_ENDIAN_UNKNOWN;
   info->osabi = GDB_OSABI_UNINITIALIZED;
+}
+
+/* Similar it init, but this time fill in the blanks.  Information is
+   obtained from the specified architecture, global "set ..." options,
+   and explicitly initialized INFO fields.  */
+
+void
+gdbarch_info_fill (struct gdbarch *gdbarch, struct gdbarch_info *info)
+{
+  /* "(gdb) set architecture ...".  */
+  if (info->bfd_arch_info == NULL
+      && !target_architecture_auto
+      && gdbarch != NULL)
+    info->bfd_arch_info = gdbarch_bfd_arch_info (gdbarch);
+  if (info->bfd_arch_info == NULL
+      && info->abfd != NULL
+      && bfd_get_arch (info->abfd) != bfd_arch_unknown
+      && bfd_get_arch (info->abfd) != bfd_arch_obscure)
+    info->bfd_arch_info = bfd_get_arch_info (info->abfd);
+  if (info->bfd_arch_info == NULL
+      && gdbarch != NULL)
+    info->bfd_arch_info = gdbarch_bfd_arch_info (gdbarch);
+
+  /* "(gdb) set byte-order ...".  */
+  if (info->byte_order == BFD_ENDIAN_UNKNOWN
+      && !target_byte_order_auto
+      && gdbarch != NULL)
+    info->byte_order = gdbarch_byte_order (gdbarch);
+  /* From the INFO struct.  */
+  if (info->byte_order == BFD_ENDIAN_UNKNOWN
+      && info->abfd != NULL)
+    info->byte_order = (bfd_big_endian (info->abfd) ? BFD_ENDIAN_BIG
+		       : bfd_little_endian (info->abfd) ? BFD_ENDIAN_LITTLE
+		       : BFD_ENDIAN_UNKNOWN);
+  /* From the current target.  */
+  if (info->byte_order == BFD_ENDIAN_UNKNOWN
+      && gdbarch != NULL)
+    info->byte_order = gdbarch_byte_order (gdbarch);
+
+  /* "(gdb) set osabi ...".  Handled by gdbarch_lookup_osabi.  */
+  if (info->osabi == GDB_OSABI_UNINITIALIZED)
+    info->osabi = gdbarch_lookup_osabi (info->abfd);
+  if (info->osabi == GDB_OSABI_UNINITIALIZED
+      && gdbarch != NULL)
+    info->osabi = gdbarch_osabi (gdbarch);
+
+  /* Must have at least filled in the architecture.  */
+  gdb_assert (info->bfd_arch_info != NULL);
 }
 
 /* */
Index: arch-utils.h
===================================================================
RCS file: /cvs/src/src/gdb/arch-utils.h,v
retrieving revision 1.60
diff -u -r1.60 arch-utils.h
--- arch-utils.h	9 Nov 2003 12:07:16 -0000	1.60
+++ arch-utils.h	9 Nov 2003 23:42:45 -0000
@@ -153,6 +153,12 @@
    default values are not zero.  */
 extern void gdbarch_info_init (struct gdbarch_info *info);
 
+/* Similar it init, but this time fill in the blanks.  Information is
+   obtained from the specified architecture, global "set ..." options,
+   and explicitly initialized INFO fields.  */
+extern void gdbarch_info_fill (struct gdbarch *gdbarch,
+			       struct gdbarch_info *info);
+
 /* Return the architecture for ABFD.  If no suitable architecture
    could be find, return NULL.  */
 
Index: gdbarch.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.c,v
retrieving revision 1.256
diff -u -r1.256 gdbarch.c
--- gdbarch.c	9 Nov 2003 12:07:16 -0000	1.256
+++ gdbarch.c	9 Nov 2003 23:43:07 -0000
@@ -5922,38 +5922,7 @@
 
   /* Fill in missing parts of the INFO struct using a number of
      sources: ``set ...''; INFOabfd supplied; existing target.  */
-
-  /* ``(gdb) set architecture ...'' */
-  if (info.bfd_arch_info == NULL
-      && !TARGET_ARCHITECTURE_AUTO)
-    info.bfd_arch_info = TARGET_ARCHITECTURE;
-  if (info.bfd_arch_info == NULL
-      && info.abfd != NULL
-      && bfd_get_arch (info.abfd) != bfd_arch_unknown
-      && bfd_get_arch (info.abfd) != bfd_arch_obscure)
-    info.bfd_arch_info = bfd_get_arch_info (info.abfd);
-  if (info.bfd_arch_info == NULL)
-    info.bfd_arch_info = TARGET_ARCHITECTURE;
-
-  /* ``(gdb) set byte-order ...'' */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN
-      && !TARGET_BYTE_ORDER_AUTO)
-    info.byte_order = TARGET_BYTE_ORDER;
-  /* From the INFO struct. */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN
-      && info.abfd != NULL)
-    info.byte_order = (bfd_big_endian (info.abfd) ? BFD_ENDIAN_BIG
-		       : bfd_little_endian (info.abfd) ? BFD_ENDIAN_LITTLE
-		       : BFD_ENDIAN_UNKNOWN);
-  /* From the current target. */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN)
-    info.byte_order = TARGET_BYTE_ORDER;
-
-  /* ``(gdb) set osabi ...'' is handled by gdbarch_lookup_osabi.  */
-  if (info.osabi == GDB_OSABI_UNINITIALIZED)
-    info.osabi = gdbarch_lookup_osabi (info.abfd);
-  if (info.osabi == GDB_OSABI_UNINITIALIZED)
-    info.osabi = current_gdbarch->osabi;
+  gdbarch_info_fill (current_gdbarch, &info);
 
   /* Must have found some sort of architecture. */
   gdb_assert (info.bfd_arch_info != NULL);
Index: gdbarch.sh
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.sh,v
retrieving revision 1.281
diff -u -r1.281 gdbarch.sh
--- gdbarch.sh	9 Nov 2003 12:07:16 -0000	1.281
+++ gdbarch.sh	9 Nov 2003 23:43:30 -0000
@@ -2138,38 +2138,7 @@
 
   /* Fill in missing parts of the INFO struct using a number of
      sources: \`\`set ...''; INFOabfd supplied; existing target.  */
-
-  /* \`\`(gdb) set architecture ...'' */
-  if (info.bfd_arch_info == NULL
-      && !TARGET_ARCHITECTURE_AUTO)
-    info.bfd_arch_info = TARGET_ARCHITECTURE;
-  if (info.bfd_arch_info == NULL
-      && info.abfd != NULL
-      && bfd_get_arch (info.abfd) != bfd_arch_unknown
-      && bfd_get_arch (info.abfd) != bfd_arch_obscure)
-    info.bfd_arch_info = bfd_get_arch_info (info.abfd);
-  if (info.bfd_arch_info == NULL)
-    info.bfd_arch_info = TARGET_ARCHITECTURE;
-
-  /* \`\`(gdb) set byte-order ...'' */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN
-      && !TARGET_BYTE_ORDER_AUTO)
-    info.byte_order = TARGET_BYTE_ORDER;
-  /* From the INFO struct. */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN
-      && info.abfd != NULL)
-    info.byte_order = (bfd_big_endian (info.abfd) ? BFD_ENDIAN_BIG
-		       : bfd_little_endian (info.abfd) ? BFD_ENDIAN_LITTLE
-		       : BFD_ENDIAN_UNKNOWN);
-  /* From the current target. */
-  if (info.byte_order == BFD_ENDIAN_UNKNOWN)
-    info.byte_order = TARGET_BYTE_ORDER;
-
-  /* \`\`(gdb) set osabi ...'' is handled by gdbarch_lookup_osabi.  */
-  if (info.osabi == GDB_OSABI_UNINITIALIZED)
-    info.osabi = gdbarch_lookup_osabi (info.abfd);
-  if (info.osabi == GDB_OSABI_UNINITIALIZED)
-    info.osabi = current_gdbarch->osabi;
+  gdbarch_info_fill (current_gdbarch, &info);
 
   /* Must have found some sort of architecture. */
   gdb_assert (info.bfd_arch_info != NULL);

Re: [commit] New function gdbarch_infi_fill

[Daniel Jacobowitz]

On Sun, Nov 09, 2003 at 06:50:54PM -0500, Andrew Cagney wrote:
> Hello,
> 
> The attached adds the new function gdbarch_info_fill.  This new function 
> fills in un-initialized fields of the "info" structure using global 
> "set" variables and the "gdbarch" parameter.
> 
> committed,
> Andrew

Yeah!  I've wanted this before.  Though now I can't remember why.

> +/* Similar it init, but this time fill in the blanks.  Information is

"Similar to".

> +/* Similar it init, but this time fill in the blanks.  Information is

Ditto.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [commit] New function gdbarch_infi_fill

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 728 bytes --]

> On Sun, Nov 09, 2003 at 06:50:54PM -0500, Andrew Cagney wrote:
> 
>> Hello,
>> 
>> The attached adds the new function gdbarch_info_fill.  This new function 
>> fills in un-initialized fields of the "info" structure using global 
>> "set" variables and the "gdbarch" parameter.
>> 
>> committed,
>> Andrew
> 
> 
> Yeah!  I've wanted this before.  Though now I can't remember why.

This all begs two questions.

Should gdbarch_info be filled in first?

Should the assertion:

	gdbarch_find_by_info (gdbarch_info (gdbarch)) == gdbarch

always hold?  At present architectures like the MIPS make this very hard 
  - keying off information found in the BFD and not the corresponding 
info struct.


> "Similar to".

Fixed.

Andrew


[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 1424 bytes --]

2003-11-10  Andrew Cagney  <cagney@redhat.com>

	* arch-utils.h, arch-utils.c: Fix typo in comment.

Index: arch-utils.c
===================================================================
RCS file: /cvs/src/src/gdb/arch-utils.c,v
retrieving revision 1.105
diff -u -r1.105 arch-utils.c
--- arch-utils.c	10 Nov 2003 16:03:26 -0000	1.105
+++ arch-utils.c	10 Nov 2003 16:16:22 -0000
@@ -686,7 +686,7 @@
   info->osabi = GDB_OSABI_UNINITIALIZED;
 }
 
-/* Similar it init, but this time fill in the blanks.  Information is
+/* Similar to init, but this time fill in the blanks.  Information is
    obtained from the specified architecture, global "set ..." options,
    and explicitly initialized INFO fields.  */
 
Index: arch-utils.h
===================================================================
RCS file: /cvs/src/src/gdb/arch-utils.h,v
retrieving revision 1.61
diff -u -r1.61 arch-utils.h
--- arch-utils.h	9 Nov 2003 23:52:28 -0000	1.61
+++ arch-utils.h	10 Nov 2003 16:16:22 -0000
@@ -153,7 +153,7 @@
    default values are not zero.  */
 extern void gdbarch_info_init (struct gdbarch_info *info);
 
-/* Similar it init, but this time fill in the blanks.  Information is
+/* Similar to init, but this time fill in the blanks.  Information is
    obtained from the specified architecture, global "set ..." options,
    and explicitly initialized INFO fields.  */
 extern void gdbarch_info_fill (struct gdbarch *gdbarch,

Re: [commit] New function gdbarch_infi_fill

[Daniel Jacobowitz]

On Mon, Nov 10, 2003 at 11:19:05AM -0500, Andrew Cagney wrote:
> >On Sun, Nov 09, 2003 at 06:50:54PM -0500, Andrew Cagney wrote:
> >
> >>Hello,
> >>
> >>The attached adds the new function gdbarch_info_fill.  This new function 
> >>fills in un-initialized fields of the "info" structure using global 
> >>"set" variables and the "gdbarch" parameter.
> >>
> >>committed,
> >>Andrew
> >
> >
> >Yeah!  I've wanted this before.  Though now I can't remember why.
> 
> This all begs two questions.
> 
> Should gdbarch_info be filled in first?

Not sure.

> Should the assertion:
> 
> 	gdbarch_find_by_info (gdbarch_info (gdbarch)) == gdbarch
> 
> always hold?  At present architectures like the MIPS make this very hard 
>  - keying off information found in the BFD and not the corresponding 
> info struct.

This, however, I'd really like.  I remember why I wanted it now.  Take
a look at the end of osabi.c:set_osabi - I am not especially confident
that that code works.  It seems to me that the above assertion makes it
much more likely to work.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [commit] New function gdbarch_infi_fill

[Andrew Cagney]

> Should the assertion:
>> 
>> 	gdbarch_find_by_info (gdbarch_info (gdbarch)) == gdbarch
>> 
>> always hold?  At present architectures like the MIPS make this very hard 
>>  - keying off information found in the BFD and not the corresponding 
>> info struct.
> 
> 
> This, however, I'd really like.  I remember why I wanted it now.  Take
> a look at the end of osabi.c:set_osabi - I am not especially confident
> that that code works.  It seems to me that the above assertion makes it
> much more likely to work.

Like with "frame_id", it would mean adding some sort of "special" field. 
  That way the MIPS could record that its debugging "o64" but with 
variation #28: sizeof(long) == 8, FPU :-/

It would also make architecture creation more robust vis:

- fill in "gdbarch_info"
Setting osabi et.al. based on the bfd et.al.

- create the architecture from "gdbarch_info"
but with _no_ bfd.

harder to get right, but probably more correct.

Adding to wish list ...

Andrew