[PATCH ARC 3/8] GAS: new ARC port

[Nick Clifton]

Hi Claudiu,

> I've made a patch for the gas' testsuite folder. It updates all the tests to the newest ARC ISA standard. Please remark that bic.d, extensions.d and warn.s are failing due to some missing features that are not in the proposed ARC port yet.

Thanks very much.  With this patch in place I am now satisfied with the 
submission over all and so I have checked it in.

There is one problem however.  I am seeing several failures in the 
binutils tests for the ARC target:

   GAS REGRESSION: DWARF2 1
   GAS REGRESSION: DWARF2 2
   GAS REGRESSION: DWARF2 3
   GAS REGRESSION: lns-duplicate
   GAS REGRESSION: lns-common-1
   LD REGRESSION: ld-discard/zero-range
   LD REGRESSION: ld-discard/zero-rel
   LD REGRESSION: ld-elf/linkonce1
   BIN REGRESSION: objdump compress debug sections
   BIN REGRESSION: objdump compress debug sections 3
   BIN REGRESSION: objcopy compress debug sections in
   BIN REGRESSION: objdump compress debug sections 3
   BIN REGRESSION: objdump -W
   BIN REGRESSION: objdump -WL

I think that these are all due to just one problem - the failure of the 
assembler to generate properly formatted DWARF debugging information for 
the ARC target.  Looking at the first gas failure for example:

   regexp_diff match failure
   regexp "^    <3c>   DW_AT_type        : <0x4a>$"
   line   "    <3c>   DW_AT_type        : <0x4a0000>"
   FAIL: DWARF2 1

To me this looks like an endian issue.  Ie it appears as if the ARC port 
is using a big endian relocation to generate what should be a little 
endian value.

Please could you investigate this problem, and if possible, fix it.

Cheers
   Nick