Commit: Improve readelf's decoding or ARM unwind tables.

[nick clifton]

Hi Hans-Peter,

> This caused, for cris-elf and cris-linux:
>
> Running /tmp/hpautotest-binutils/bsrc/src/ld/testsuite/ld-cris/cris.exp ...
> FAIL: ld-cris/tls-js1
> FAIL: ld-cris/tls-local-63
> FAIL: ld-cris/tls-local-64
> FAIL: ld-cris/tls-und-38
> FAIL: ld-cris/tls-und-42
> FAIL: ld-cris/tls-und-46
> FAIL: ld-cris/tls-und-50
> FAIL: ld-cris/weakref3
> FAIL: ld-cris/weakref4

Oops - sorry.  I have checked in a patch to fix this.

You may feel however that I have chosen the wrong fix.  The tests were 
expecting readelf to "there are no unwind sections in this file" whereas 
I changed it to the more accurate "readelf does not know how to decode 
unwind sections for this architecture".  Ie there *might* be unwind 
sections in the file, but readelf does not know what they look like (or 
their names) so it will not decode them.

Cheers
   Nick

ld/testsuite/ChangeLog
2011-12-05  Nick Clifton  <nickc@redhat.com>

	* ld-cris/weakref4.d: Updated pattern match for the message from
	readelf stating that there are no decoded unwind sections.
	* tls-js1.d: Likewise.
	* tls-local-63.d: Likewise.
	* tls-local-64.d: Likewise.
	* tls-und-38.d: Likewise.
	* tls-und-42.d: Likewise.
	* tls-und-46.d: Likewise.
	* tls-und-50.d: Likewise.
	* weakref3.d: Likewise.
	* weakref4.d: Likewise.