inline (and variants) function modifier

inline (and variants) function modifier

[Jan Beulich]

May I ask how one is supposed to write a non-static function and let the
compiler decide whether it is worth inlining? Since all of 'inline',
'__inline', and '__inline__' get __attribute__((always_inline))
attached, I can't see how I would currently do this. Wouldn't it make
sense to leave at least one of the three with its original meaning?

Additionally, while on a subject like this, is there a reason
attributes (as the above) are generally specified without leading (and
trailing) double underscores? This way, if I #define a symbol with the
name of an existing (or even future) attribute, things are going to
break, whereas conventions preclude me from #define-ing symbols with
double leading underscores.

Thank you,
Jan

Re: inline (and variants) function modifier

[Jan Engelhardt]

>May I ask how one is supposed to write a non-static function and let the
>compiler decide whether it is worth inlining? Since all of 'inline',
>'__inline', and '__inline__' get __attribute__((always_inline))
>attached, I can't see how I would currently do this. Wouldn't it make
>sense to leave at least one of the three with its original meaning?

The only way I see ATM is to:
#undef inline

Then inline and __attribute__((always_inline)) are "independent" again.



Jan Engelhardt
-- 
Gesellschaft für Wissenschaftliche Datenverarbeitung
Am Fassberg, 37077 Göttingen, www.gwdg.de