x86: remove i386-opc.c

[Alan Modra]

On Thu, Dec 22, 2022 at 09:49:41AM +0100, Jan Beulich wrote:
> On 20.12.2022 12:39, Alan Modra wrote:
> > On Tue, Dec 20, 2022 at 09:40:14AM +0100, Jan Beulich wrote:
> >> On 20.12.2022 09:33, Alan Modra via Binutils wrote:
> >>> Regen opcodes/po/POTFILES.in
> >>>
> >>> diff --git a/opcodes/po/POTFILES.in b/opcodes/po/POTFILES.in
> >>> index 9906bf14848..effd73f8f78 100644
> >>> --- a/opcodes/po/POTFILES.in
> >>> +++ b/opcodes/po/POTFILES.in
> >>> @@ -73,7 +73,6 @@ h8300-dis.c
> >>>  hppa-dis.c
> >>>  i386-dis.c
> >>>  i386-gen.c
> >>> -i386-opc.c
> >>>  i386-opc.h
> >>>  ia64-asmtab.c
> >>>  ia64-asmtab.h
> >>
> >> Which makes me wonder: If this is the list of files where string literals
> >> may/will need translation, are files like i386-gen.c (i.e. build time
> >> helper tool sources) really supposed to be enumerated here?
> > 
> > Probably not, but the build time tools don't tend to use gettext and
> > thus all it does is slow down generation of .pot files a little.
> 
> Irrespective of them using gettext (I've now queued a patch to remove
> that from i386-gen.c), wouldn't it make sense to convert
> 
> POTFILES = $(HFILES) $(CFILES)
> 
> to
> 
> POTFILES = $(HFILES) $(LIBOPCODES_CFILES)
> 
> thus eliminating translation for build time tools regardless? At which
> point the question would be whether CFILES as a variable needs retaining.
> Afaics the definition of POTFILES is the only user.

Sounds reasonable to me.

-- 
Alan Modra
Australia Development Lab, IBM