Explicit architecture for m68k tests

Explicit architecture for m68k tests

[Vincent Rivière]

[-- Attachment #1: Type: text/plain, Size: 333 bytes --]

Hello.

Due to the limitations of the a.out format, the target architecture can't be 
automatically guessed by objdump.

The attached patch is required to make all tests pass on m68k-aout targets. 
It explicitly sets the architecture for objdump.
Please apply to the 2.21 branch as well as the trunk.

Regards,

-- 
Vincent Rivière

[-- Attachment #2: m68k-arch.patch --]
[-- Type: text/plain, Size: 1486 bytes --]

diff -aurN binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/cpu32.d binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/cpu32.d
--- binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/cpu32.d	2010-11-20 12:31:53.093750000 +0100
+++ binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/cpu32.d	2010-11-21 15:00:27.578125000 +0100
@@ -1,5 +1,5 @@
 #name: cpu32
-#objdump: -d
+#objdump: --architecture=m68k:cpu32 -d
 #as: -mcpu32
 
 .*:     file format .*
diff -aurN binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/mcf-coproc.d binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/mcf-coproc.d
--- binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/mcf-coproc.d	2010-11-20 12:31:53.125000000 +0100
+++ binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/mcf-coproc.d	2010-11-21 14:55:24.750000000 +0100
@@ -1,4 +1,4 @@
-#objdump: -d
+#objdump: --architecture=m68k:547x -d
 #as: -mcpu=5475
 
 .*:     file format .*
diff -aurN binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/mcf-wdebug.d binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/mcf-wdebug.d
--- binutils-CVS-2_21-branch-20101120/gas/testsuite/gas/m68k/mcf-wdebug.d	2010-11-20 12:31:53.140625000 +0100
+++ binutils-CVS-2_21-branch-20101120-mint/gas/testsuite/gas/m68k/mcf-wdebug.d	2010-11-21 16:10:02.515625000 +0100
@@ -1,5 +1,5 @@
 #name: mcf-wdebug
-#objdump: -d
+#objdump: --architecture=m68k:5200 -d
 #as: -m5208
 
 .*:     file format .*

Re: Explicit architecture for m68k tests

[Alan Modra]

On Sun, Nov 21, 2010 at 04:19:16PM +0100, Vincent Rivière wrote:
> The attached patch is required to make all tests pass on m68k-aout
> targets. It explicitly sets the architecture for objdump.

I don't see any problem when I build ab m68k-aout target, even with
--enable-targets=all.  How did you configure?

-- 
Alan Modra
Australia Development Lab, IBM

Re: Explicit architecture for m68k tests

[Vincent Rivière]

Alan Modra wrote:
> I don't see any problem when I build ab m68k-aout target, even with
> --enable-targets=all.  How did you configure?

../src/configure --target=m68k-aout --disable-nls
make
cd gas
make check
FAIL: gas/m68k/mcf-coproc
FAIL: mcf-wdebug
FAIL: cpu32

With the proposed patch, the whole gas testsuite succeeds.

-- 
Vincent Rivière

Re: Explicit architecture for m68k tests

[Alan Modra]

On Mon, Nov 22, 2010 at 06:44:48PM +0100, Vincent Rivière wrote:
> Alan Modra wrote:
> >I don't see any problem when I build an m68k-aout target, even with
> >--enable-targets=all.  How did you configure?
> 
> ../src/configure --target=m68k-aout --disable-nls
> make
> cd gas
> make check
> FAIL: gas/m68k/mcf-coproc
> FAIL: mcf-wdebug
> FAIL: cpu32
> 
> With the proposed patch, the whole gas testsuite succeeds.

I thought m68k-netbsd was aout, and running the testsuite on that
target succeeds.  Huh, mystery solved.  The difference is that
m68k-aout has a default target vector of aout0_big_vec while
m68k-netbsd uses m68knetbsd_vec.  The latter has the architecture, the
former doesn't.

Patch OK.

-- 
Alan Modra
Australia Development Lab, IBM

Re: Explicit architecture for m68k tests

[Vincent Rivière]

Le 22/11/2010 22:36, Alan Modra a écrit :
>> ../src/configure --target=m68k-aout --disable-nls
>> make
>> cd gas
>> make check
>> FAIL: gas/m68k/mcf-coproc
>> FAIL: mcf-wdebug
>> FAIL: cpu32
>>
>> With the proposed patch, the whole gas testsuite succeeds.
>
> I thought m68k-netbsd was aout, and running the testsuite on that
> target succeeds.  Huh, mystery solved.  The difference is that
> m68k-aout has a default target vector of aout0_big_vec while
> m68k-netbsd uses m68knetbsd_vec.  The latter has the architecture, the
> former doesn't.
>
> Patch OK.

Alan, could you please commit my patch ?

For memory:
http://sourceware.org/ml/binutils/2010-11/msg00384.html

-- 
Vincent Rivière

Re: Explicit architecture for m68k tests

[Alan Modra]

On Wed, Feb 09, 2011 at 10:48:34PM +0100, Vincent Rivière wrote:
> Le 22/11/2010 22:36, Alan Modra a écrit :
> >Patch OK.
> 
> Alan, could you please commit my patch ?
> 
> For memory:
> http://sourceware.org/ml/binutils/2010-11/msg00384.html

Applied.

-- 
Alan Modra
Australia Development Lab, IBM

Re: Explicit architecture for m68k tests

[Vincent Rivière]

Alan Modra wrote:
> Applied.

Good.
Now the gas testsuite for m68k-aout targets success.

-- 
Vincent Rivière