> However, I am still confused by the fact that the %eax is not used by
> mcount(), and `gcc -pg' generated code emits .long 0's, but doesn't
> reference them at all.  
> 
> It's probably just an anachronism.  Probably gcc should be changed to
> stop emitting the .long 0's; and the ENTRY() macro in <machine/asm.h>
> should be changed to omit the .long 0 and the movl $1b,%eax.
> 

I removed the reference to it from the i386 tm.h's FUNCTION_PROFILER() macro.
The .long 0's itself are emitted from common (ie. non-i386 specific) code, in `final.c',
which I didn't want to touch at the time without inspecting other ports.

These .long's, someone told me some time ago, may have been in use in the days of prof(1).

-pk

------------------------------------------------------------------------------