RFA: Exhume the ARM port [PR icedtea/484, 323]

[Andrew Haley]

On 10/24/2011 11:32 AM, Matthias Klose wrote:
> On 10/20/2011 03:55 PM, Andrew Haley wrote:
>> 	* openjdk-ecj/hotspot/src/cpu/zero/vm/cppInterpreter_arm.S
>> 	(REWRITE_PAIRS): New macro.
>> 	* openjdk-ecj/hotspot/src/cpu/zero/vm/bytecodes_arm.def: Use
>> 	REWRITE_PAIRS to prevent rewriting pairs of bytecodes in the
>> 	instruction stream.
> 
> this should be
> arm_port/hotspot/src/cpu/zero/vm/cppInterpreter_arm.S
> arm_port/hotspot/src/cpu/zero/vm/bytecodes_arm.def
> 
> test results for build with -mfloat-abi=softfp and mfloat-abi=hard are attached.
> The build fails when building with -march=armv7a -mthumb.

I'm not at all surprised.  I think we should simply mark this
combination unsupported for the time being.

> LD_LIBRARY_PATH=. ./gamma -Xbatch -showversion Queens
> Illegal instruction
> 
> Starting program:
> /scratch/packages/openjdk/b23/openjdk-6-6b23~pre11/build/openjdk.build/hotspot/outputdir/linux_arm_zero/product/gamma
> -Xbatch -showversion Queens
> [Thread debugging using libthread_db enabled]
> 
> Program received signal SIGILL, Illegal instruction.
> 0x400faef2 in Thumb2_Clear_Cache () from ./libjvm.so

I'll see if I can duplicate this.

> I get a successful build when adding -marm to CFLAGS in
> hotspot/make/linux/makefiles/gcc.make (having a compiler configured with
> --with-arch=armv7-a --with-float=softfp --with-fpu=vfpv3-d16 --with-mode=thumb)
> 
> the jdk tests run up to
> Passed: java/beans/PropertyChangeSupport/Test4092905.java
> Passed: java/beans/PropertyChangeSupport/Test4353056.java
> 
> then terminate. no error message.

OK, so I think we're good enough for this to go into trunk, for
more testing.

Andrew.