coff testsuite failures

[Nick Clifton]

Hi Hans-Peter,

> I tested on cris-axis-elf, i686-pc-linux-gnu and
> mmix-knuth-mmixware which I thought would represent all
> necessary flavors.  Now I know better.

My complete testsuite checks 69 different targets which is probably
overkill but it never hurts to be sure.

>>   alpha-unknown-osf4.0  : Reports the symbols as having type 'N' not 't'.
>
> Seems like a bug; .text isn't a debug section, if this 'N' is as
> in bfd/syms.c.

Yup - in fact I felt that all three targets which currently fail do so
because of real bugs, so I have not tried to fix the testsuite to
accommodate them or turn them into XFAILS.  (I would like to see them
fixed).


> Does anyone have a favourite list of target to test sweeping
> changes on?  I'm finished with my "ld" stuff for MMIX and would
> like to test it broadly.  (Or I'll become desperate and look in
> ld/configure.tgt and the mailing list archive. :-)

I would recommend at least the following:

  x86 native         [Other natives if you have access to them.
                      Cygwin and Sparc would be good]

  all-64             [ie configured with --enable-targets=all
                      --enable-64-bit-bfd] 

  arm-elf            [Well it is my favourite embedded part]
  
  mips-elf           [Lots of people like the MIPS]
  
  z8k-coff           [Or a.n.other COFF based port]

  mingw32-pe         [Or a.n.other PE based port]
    
  rs6000-aix4.3.3    [If you do not have access to native hosts, at
                      least make sure that the cross compilers work.
                      I would suggest adding alpha, cygwin, hppa and
                      sparc to this list as well]

  sparc64-linux-gnu  [At least one 64-bit part where you can check the
                      assembler is a good idea as well].

Cheers
        Nick