Re-RFA re: [almost-obvious patch, and 6.0 RFA] treat mipsisa64 as 64-bit MIPS.

Re-RFA re: [almost-obvious patch, and 6.0 RFA] treat mipsisa64 as 64-bit MIPS.

[cgd]

[-- Attachment #1: Type: text/plain, Size: 86 bytes --]

been over a week, normally wouldn't resend it, but Andrew's msg
spurred me on.  8-)



[-- Attachment #2: Type: message/rfc822, Size: 7163 bytes --]

From: cgd@broadcom.com
To: gdb-patches@sources.redhat.com
Subject: [almost-obvious patch, and 6.0 RFA] treat mipsisa64 as 64-bit MIPS.
Date: Wed, 13 Aug 2003 19:43:49 +0000 (UTC)
Message-ID: <mailpost.1060803829.14185@news-sj1-1>

Currently, the gdb configury doesn't DTRT with "mipsisa64", i.e., it
doesn't treat it as 64-bit MIPS.

This causes problems when you want to debug with sim, you end with the
error:

18 [ldt-sj3-010] bin % mipsisa64-elf-gdb 
tGNU gdb 2003-08-11-cvs
Copyright 2003 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "--host=i686-pc-linux-gnu --target=mipsisa64-elf".a
(gdb) target sim
Size of register  (0/0) incorrect (8 instead of 4))Connected to the simulator.
(gdb) 

(I'm sure it causes other problems, too.  8-)


So, this patch adds 'mipsisa64' variants for all of the generic
'mips64' configure.tgt entries.

The reason this patch is *almost* but not quite obvious is that I
don't know what the mips64*-big-* target is about.  It looks suitably
generic that, assuming it should exist at all, it should have a
mipsisa64* entry.  However i really don't know.  (The -big- and
-little- targets are confusing to me.  I mean, why do they exist any
more at all, and why is there only a 32-bit -little- target and no
64-bit one?)

Note that I don't think that most of these targets are horribly useful
anymore, but they should probably have "mipsisa64" variants for
completeness.

(The issue is that if you configure a mipsisa64<some-cpu>-... GDB,
e.g. mipsisa64sb1-..., you should get different simulator bits than if
you do mips64<some-other-cpu>-...  Personally, the more I think about
it, I think that the choice of "mipsisa32" and "mipsisa64" to name
MIPS32/MIPS64 ISA targets was unfortunate...  But I didn't set the
trend, and it would have been wrong to name MIPS64 something other
than mipsisa64 once mipsisa32 existed...  *sigh*)


Verified that this does the expected thing on the trunk with
mipsisa64-elf.

Also verified that it applies cleanly to the 6.0-branch configure.tgt,
but not tested further than that.  (read it, it won't cause harm.  8-)




cgd
--
2003-08-13  Chris Demetriou  <cgd@broadcom.com>

        * configure.tgt (mipsisa64*-big-*, mipsisa64*el-*-ecoff*)
        (mipsisa64*-*-ecoff*, mipsisa64*el-*-elf*)
        (mipsisa64*-*-elf*, mipsisa64*-*-linux*): Treat identically
        to corresponding "mips64" configurations.

Index: configure.tgt
===================================================================
RCS file: /cvs/src/src/gdb/configure.tgt,v
retrieving revision 1.111
diff -u -p -r1.111 configure.tgt
--- configure.tgt	1 Aug 2003 21:14:33 -0000	1.111
+++ configure.tgt	13 Aug 2003 17:43:12 -0000
@@ -144,16 +144,22 @@ m68*-*-vxworks*)	gdb_target=vxworks68 ;;
 
 mcore*-*-*)		gdb_target=mcore ;;
 mips64*-big-*)		gdb_target=mips64 ;;
+mipsisa64*-big-*)	gdb_target=mips64 ;;
 mips*-big-*)		gdb_target=mips ;;
 mips*-dec-*)		gdb_target=mips ;;
 mips*-*-pe)		gdb_target=wince ;;
 mips64*el-*-ecoff*)	gdb_target=embedl64 ;;
+mipsisa64*el-*-ecoff*)	gdb_target=embedl64 ;;
 mips64*-*-ecoff*)	gdb_target=embed64 ;;
+mipsisa64*-*-ecoff*)	gdb_target=embed64 ;;
 mips64*vr*-*-elf*)	gdb_target=mips64 ;;
 mips*tx39*-elf*)	gdb_target=tx39 ;;
 mips64*el-*-elf*)	gdb_target=embedl64 ;;
+mipsisa64*el-*-elf*)	gdb_target=embedl64 ;;
 mips64*-*-elf*)		gdb_target=embed64 ;;
+mipsisa64*-*-elf*)	gdb_target=embed64 ;;
 mips64*-*-linux*)	gdb_target=linux64 ;;
+mipsisa64*-*-linux*)	gdb_target=linux64 ;;
 mips*el-*-ecoff*)	gdb_target=embedl ;;
 mips*-*-ecoff*)		gdb_target=embed ;;
 mips*el-*-elf*)		gdb_target=embedl ;;


[-- Attachment #3: Type: text/plain, Size: 229 bytes --]



-- 
Chris Demetriou                                            Broadcom Corporation
                Principal Design Engineer, Broadband Processors
  Any opinions expressed in this message are mine, not necessarily Broadcom's.

Re: Re-RFA re: [almost-obvious patch, and 6.0 RFA] treat mipsisa64 as 64-bit MIPS.

[Andrew Cagney]

> So, this patch adds 'mipsisa64' variants for all of the generic
> 'mips64' configure.tgt entries.

Ah.

> Verified that this does the expected thing on the trunk with
> mipsisa64-elf.
> 
> Also verified that it applies cleanly to the 6.0-branch configure.tgt,
> but not tested further than that.  (read it, it won't cause harm.  8-)

Yes, for both, but a few tweaks:

- put the patterns at the start, vis:
	mipsisa64*-*-linux* ) gdb_target=linux64 ;;
	mipsisa64*-*-* ) gdb_target=embed64 ;;
I think this is all that is needed.  It seems that my recent MIPS target 
purge missed a few cases.

- add brief comment alerting the user as to the issue.

It can probably be simplified further (mips*64*-*-*?) but, as you note, 
harm minimization is in order.

Andrew


> 2003-08-13  Chris Demetriou  <cgd@broadcom.com>
> 
>         * configure.tgt (mipsisa64*-big-*, mipsisa64*el-*-ecoff*)
>         (mipsisa64*-*-ecoff*, mipsisa64*el-*-elf*)
>         (mipsisa64*-*-elf*, mipsisa64*-*-linux*): Treat identically
>         to corresponding "mips64" configurations.
> 

Re: Re-RFA re: [almost-obvious patch, and 6.0 RFA] treat mipsisa64 as 64-bit MIPS.

[cgd]

At Mon, 25 Aug 2003 15:30:47 +0000 (UTC), "Andrew Cagney" wrote:
> - put the patterns at the start, vis:
> 	mipsisa64*-*-linux* ) gdb_target=linux64 ;;
> 	mipsisa64*-*-* ) gdb_target=embed64 ;;
> I think this is all that is needed.  It seems that my recent MIPS
> target purge missed a few cases.

ahh, yes.

Looks like:

        tx39.mt == tx39l.mt
        embed64.mt == embedl64.mt
        mips.mt == embedl.mt (!!! except for comments; this one is puzzling!)

so probably more could be collapsed later.


> It can probably be simplified further (mips*64*-*-*?) but, as you
> note, harm minimization is in order.

Yeah, for some reason i don't like using "mips*64".  dunno why it bugs
me, though.


Patch as re-tested with mipsisa64-elf gdb and checked in (trunk &
branch) is below.


thanks,

chris
--
2003-08-25  Chris Demetriou  <cgd@broadcom.com>

	* configure.tgt: Document need for special "mipsisa64" handling.
	(mipsisa64*-*-linux64): Handle as target linux64.
	(mipsisa64*-*-*): Handle as target embed64.

Index: configure.tgt
===================================================================
RCS file: /cvs/src/src/gdb/configure.tgt,v
retrieving revision 1.111
diff -u -p -r1.111 configure.tgt
--- configure.tgt	1 Aug 2003 21:14:33 -0000	1.111
+++ configure.tgt	25 Aug 2003 22:55:39 -0000
@@ -143,6 +143,12 @@ m68*-*-sysv4*)		gdb_target=m68kv4 ;;
 m68*-*-vxworks*)	gdb_target=vxworks68 ;;
 
 mcore*-*-*)		gdb_target=mcore ;;
+
+# "mipsisa64" targets are 64-bit MIPS, MIPS64 ISA.  They need these
+# special cases because otherwise they (obviously) are not matched by
+# the "mips64" target cases, and would therefore be treated as 32-bit.
+mipsisa64*-*-linux*)	gdb_target=linux64 ;;
+mipsisa64*-*-*)		gdb_target=embed64 ;;
 mips64*-big-*)		gdb_target=mips64 ;;
 mips*-big-*)		gdb_target=mips ;;
 mips*-dec-*)		gdb_target=mips ;;