Re: [patch] Associate a name with a CODE_LABEL

[Richard Earnshaw <rearnsha at arm dot com>]

> On Thu, Nov 04, 1999 at 04:51:24AM -0700, Jeffrey A Law wrote:
> >   In message <199911041058.KAA11748@cam-mail1.cambridge.arm.com>you write:
> >   > 
> >   > clm@cygnus.com said:
> >   >   + /* Provide default for ASM_OUTPUT_ALTERNATE_LABEL_NAME.  */
> >   >   + #ifndef ASM_OUTPUT_ALTERNATE_LABEL_NAME
> >   >   + #define ASM_OUTPUT_ALTERNATE_LABEL_NAME(FILE,INSN) \
> >   >   +   fprintf (FILE, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
> >   >   + #endif
> >   >   + 
> >   > 
> >   > This on its own will mean that any port that does not use ':' as a label 
> >   > declaration will now be broken (the arm-aof port springs to mind).  Would 
> >   > it not be better to define this in terms of some other label_generation 
> >   > macro (I'm not entirely sure of the intended use, but ASM_OUTPUT_LABEL 
> >   > might be more appropriate).
> > Agreed.  FWIW, the PA is another target that does not require colons (in fact
> > if you use the HP assembler, you'll get an error if you put a colon after a
> > label. ]
> 
> However since those ports don't create alternate labels, I don't see how they
> are broken.  Yes, if they create alternate labels, they have to define the
> above macro.....

You miss my point.

Almost all ports have a single fundamental way of declaring a label; 
normally, but not always, it is to put a colon after the label's name.  
What we should really have is something like ASM_OUTPUT_LABEL_DECL(char 
*label) which does this most basic operation.  Then, when we add new 
routines such as ASM_OUTPUT_ALTERNATE_LABEL_NAME, we can define the 
default in terms of the most basic macros and stand a much better chance 
of them working on all ports.

Richard.