On Wed, Nov 20, 2019 at 01:36:57PM +0000, Ramsay Jones wrote:
>
>
> On 20/11/2019 02:13, Ramsay Jones wrote:
> [snip]
>
> >
> > Hmm, shouldn't these:
> >
> >> + {"const", NS_KEYWORD, MOD_PURE, .op = &attr_fun_op },
> >> + {"__const", NS_KEYWORD, MOD_PURE, .op = &attr_fun_op },
> >> + {"__const__", NS_KEYWORD, MOD_PURE, .op = &attr_fun_op },
> >
> > ... be attr_mod_op? (I'm just reading this in my email client, so I
> > haven't given it much thought, but it just seems wrong ...)
>
> Of course, as soon as my head hit the pillow, I remembered that
> gcc has a 'const function attribute', somewhat similar to 'pure'.
> (I just looked it up, because I couldn't remember what the
> difference was between 'const' and 'pure'!).
>
> So, just ignore me! :-D
Hehe, I already had the case at several times when I saw these,
thinking "eh! that cannot possible be correct" and then "ah yes,
these are not for the 'real' const, it's for the other one" ;)
-- Luc
