Serious error-causing change in new CPP's -traditional behavior

[Dave Brolley]

Richard Earnshaw wrote:

> > Zack Weinberg wrote:
> >
> > > On Thu, May 11, 2000 at 05:03:46PM +0100, Richard Earnshaw wrote:
> > > > What I would like to have is
> > > >
> > > > #define wibble(foo) mov       r0, # foo
> > > > .text
> > > > a:
> > > >       wibble(4)
> > > >
> > > > and get the output
> > > >
> > > > .text
> > > > a:
> > > >         mov r0, #4
> > > >
> > > > Unfortunately we get
> > > >
> > > >       mov r0, "4"
> > >
> > > Alas, this may not have been what you intended, but it is the behavior
> > > mandated by the standard for that #define.  Whitespace between the
> > > stringize operator and its argument (macro name) is not significant.
> > > I don't plan to change this in assembly language mode.
> >
> > This should be made to work as Richard expects for  -traditional
> >
> > Dave
> >
>
> True, but in this context we were talking about what would happen if we
> stopped using -traditional for assembly.
>
> FWIW, I don't disagree with Zack here; this is just a gotcha that must be
> watched for; with the recent patches applied, there are now several viable
> alternatives for getting the desired output.

In that case, I agree with Zack. You need to code something like:

#define HASH #    /* '#' isn't special on object-like macros */
#define LPREFIX() HASH
#define wibble (foo) mov    r0, LPREFIX()foo
wibble (4)

This will also have the benefit of allowing you to change the literal prefix
easily for different assemblers.

Dave