Inline documentation patch...

Richard Henderson rth@cygnus.com
Mon Feb 8 20:00:00 GMT 1999


On Mon, Feb 08, 1999 at 06:56:17PM -0800, Linus Torvalds wrote:
> If "extern inline" still works as a "inline if at all humanly possible",
> then I agree wholeheartedly. 

We do that now, modulo the one test that's causing the point of
contention here (whose target problem can be solved differently,
and I'm taking care of now).

Note that we do ignore function size when making the decision
if the function is explicitly marked inline.

Here are the cases we cannot inline:

  * varargs
  * uses alloca
  * nested functions
  * return structure the old-fashioned pcc way
  * variable sized return value
  * variable sized arguments
  * contains computed jump (the point of contention)
  * contains a nonlocal goto
  * arguments with destructors (could be fixed)
  * multi-register structure return values (could be fixed) 


r~


More information about the Gcc-patches mailing list