Re: A patch to fix COFF

[Nick Clifton <nickc at redhat dot com>]

Hi H.J.

> > What do you think ?
> 
> It was my first thought. But I was too lazy to change all
> 
> config/tc-alpha.c
> config/tc-arm.c
> config/tc-cris.c
> config/tc-hppa.c
> config/tc-i386.c
> config/tc-ia64.c
> config/tc-m68hc11.c
> config/tc-mn10300.c
> config/tc-ppc.c
> config/tc-sh.c
> config/tc-sparc.c
> config/tc-v850.c

Hmm yes :-)

> If you want to go this way, please check all of them and make sure
> they are all ok. I know config/tc-i386.c is not.

Sorry - what is wrong with tc-i386.c ?  I mean, what would break if we
patched it as suggested ?

> If you believe it is
> the right approach, shouldn't the definitions of dwarf2_directive_loc,
> dwarf2_directive_file, ...  be removed if BFD_ASSEMBLER is not defined?
> You will get a link-time error instead of a run-time error.

Hmm, yes this is a very good point.  I guess that it is a question of
doing the "right thing" which would involve patching lots of ports and
has the potential to introduce lots of new bugs, or to take the
"minimal change" approach, and have the dwarf2 stuff essentially be
no-ops for non BFD assemblers.

I think that you have persuaded me, although I do have one small
reservation - shouldn't the non BFD_ASSEMBLER version of
dwarf2_directive_file() always pass 0 onto s_app_file(), rather than
passing on its appfile argument ?

Cheers
        Nick