Re: Changes in GC and in pure space

[Paul Eggert <eggert@cs.ucla.edu>]

Óscar Fuentes wrote:
> At least in the C++ world, `inline' is used for putting function
> definitions on headers, not for inlining.
> 
> There is a consensus from many years ago that the compiler is expected
> to do the right thing wrt inlining irrespectively of the presence or
> ausence of the keyword

Things are a bit different in C. In C, the consensus you mention holds for 
'static inline' (where the 'inline' is typically just a noise word nowadays), 
but it does not hold for the plain 'inline' functions that we're talking about 
because C requires that if one compilation unit defines a plain 'inline' 
function, then exactly one other compilation unit must define that function as 
'extern inline'.

Emacs rarely uses 'static inline', due to the consensus that you mention. In the 
few exceptions where Emacs does use 'static inline', I vaguely recall there 
being performance advantages when compiled by GCC in typical platforms, as GCC 
does not ignore the 'inline' in these exceptional situations.