On Sun, 06 Jul 2003, Robert Collins wrote:
> #define DoDebug(CONTENT) { \
> Debug::getDebugOut() << CONTENT; \
> Debug::finishDebug(); }
> #define debugs(SECTION, LEVEL) \
> if ((Debug::level = (LEVEL)) > Debug::Levels[SECTION]) \
> (void) 0; \
> else \
> DoDebug
I would wrap the DoDebug macro definition in the conventional
"do { ... } while (0)" construct, so that trailing semicolons work
as expected. The debugs macro can't be handled this way.
#define DoDebug(CONTENT) \
do { \
Debug::getDebugOut() << CONTENT; \
Debug::finishDebug(); \
} while (/*CONSTCOND*/ 0)
#define debugs(SECTION, LEVEL) \
if ((Debug::level = (LEVEL)) <= Debug::Levels[SECTION]) \
(void) 0;
else \
DoDebug
> /* The goal is:
> * debugs(x,y)(a << b << c)
> * to evaluate to
> * if ((Debug::level = (LEVEL)) > Debug::Levels[SECTION])
> * (void) 0;
> * else
> * { Debug::getDebugOut() << a << b << c; Debug::finishDebug();}
> */
My way, the caller must supply a trailing semicolon in
debugs(x,y)(a << b << c);
whereas your way the caller is not allowed to supply a trailing semicolon.
Supplying the trailing semicolon is more natural; it makes use of the macro
more closely resemble use of an ordinary function.
--apb (Alan Barrett)
