dtrace discuss - Dec 2007 - cpp(1) madness

Background

I''m generating rather substantial D scripts (for DTrace), using
cpp(1).  In fact, they''re so large that DTrace can''t load them
by default.  On Solaris, I have to increase a kernel value
(dtrace_dof_maxsize).  On Leopard, that option isn''t available
(yet), so I simply limit the size of my probe sets.  Sigh.

Leopard''s dtrace also has a minor bug in handing code off to
cpp for processing.  I started using cpp directly as a way to
work around this, but soon found that it was very useful to be
able to look at the expanded code.

Finally, Leopard''s cpp allows some things that neither Solaris
nor "C Reference Manual" (Harbison and Steele, 4th ed.) accept:

  % cat t1
  #define foo(bar) printf("bar=...")
  foo(abc);

  % cpp t1
  ...
  printf("abc=...");

Although this workaround is legal:

  % cat t3
  #define foo(bar) printf("%s=...", bar)
  foo("abc");

  % cpp t3
  ...
  printf("%s=...", "abc");

it makes my D scripts bigger, so the reduced size probe sets
don''t load.  So, it''a not a happy solution.

I tried messing about with concatenation macros, but so far, I''m
not getting the results I want.


Question

What (portable!) cpp syntax can I use to allow a macro parameter
to be folded into a double-quoted string?  It doesn''t have to be
pretty!

-r
-- 
http://www.cfcl.com/rdm            Rich Morin
http://www.cfcl.com/rdm/resume     rdm at cfcl.com
http://www.cfcl.com/rdm/weblog     +1 650-873-7841

Technical editing and writing, programming, and web development

>Finally, Leopard''s cpp allows some things that neither Solaris
>nor "C Reference Manual" (Harbison and Steele, 4th ed.) accept:
>  % cat t1
>  #define foo(bar) printf("bar=...")
>  foo(abc);
Yuck.  I thought Reiser CPP did that also (which is what traditional
CPP is derived from.
>  % cpp t1
>  ...
>  printf("abc=...");
>
>Although this workaround is legal:
>
>  % cat t3
>  #define foo(bar) printf("%s=...", bar)
>  foo("abc");
>
>  % cpp t3
>  ...
>  printf("%s=...", "abc");
>
>it makes my D scripts bigger, so the reduced size probe sets
>don''t load.  So, it''a not a happy solution.
>
>I tried messing about with concatenation macros, but so far, I''m
>not getting the results I want.
>
>
>Question
>
>What (portable!) cpp syntax can I use to allow a macro parameter
>to be folded into a double-quoted string?  It doesn''t have to be
>pretty!
Unfortunately, if you use "CPP" then you may not get a portable C 
preprocessor; the proper ANSI C syntax is:

#define foo(bar)	printf( #bar "=...")

However, this assumes that the underlying language can do string pasting
(in C, the syntax   "abc" "=..."  is equivalent to
"abc=...")

You can''t concatenate anything other than tokens in C (abc ## def).

#bar make a string of the value of bar, so the final expansion of foo(abc)
is

	printf("abc" "=...")

Casper