New failures from extract symbols test

[Alan Modra]

Hi Nick,
On Fri, Oct 16, 2015 at 09:13:42AM +0100, Nick Clifton wrote:
>   I am seeing some new linker testsuite failures from the extract
>   symbols test.  The targets that are failing are:
> 
>  arm-epoc-pe 
>  arm-wince-pe 
>  i586-aout 
>  i686-pc-cygwin 
>  i686-pc-mingw32 
>  mcore-pe 
>  mingw32-pe 
>  mips-elf 
>  mips64-linux 
>  mips64vrel-elf 
>  mipsel-linux-gnu 
>  mipsisa32el-linux 
>  mmix-mmixware 
>  ns32k-netbsd 
>  sh-pe 
>  tx39-elf 
>  x86_64-pc-cygwin 

Yes, I'd noticed.  I haven't looked properly into any of the failures
yet.  The i586-aout fail is just not meeting the zero size section
test.  That may not matter much.  The others all show deficiencies in
real-world use of "objcopy --extract-symbol".  We didn't see them
before due to not running any test of the option on anything but ELF
targets, and the old extract-symbol test didn't fail on MIPS due to
removing .reginfo.  When it is present we get a zero size .reginfo
which results in a warning from nm (well, bfd really) "don't know how
to handle processor specific section `.reginfo' [0x70000006]".

So objcopy hasn't gotten worse, we're just sweeping less under the
rug.

-- 
Alan Modra
Australia Development Lab, IBM