Re: [DRAGONEGG] Expose 4 functions in i386.c

[Duncan Sands <baldrick at free dot fr>]

Hi Andrew,

> Aren't at least one of these a target hook? Why can't you use the target
> hooks?

thanks for thinking about this.  As far as I can see none of these are target
hooks.  Which one were you thinking of?  As for why I can't use target hooks,
as I mentioned in my email I believe I can rework the plugin to only use target
hooks and not these functions, but it will take some time.  That's why I'm
requesting applying this for the moment.  I would particularly like to have it
in gcc-4.5.  I'd be fine with not having it in mainline, since I'm confident
that I will have resolved the issue before gcc-4.6 comes out, but I understand
that (when feasible) the policy is to not apply patches to gcc-4.5 unless they
have been applied to the mainline branch.

Ciao,

Duncan.

> Sent from my iPhone
>
> On Apr 13, 2010, at 8:31 AM, Duncan Sands <baldrick@free.fr> wrote:
>
> > My plugin currently needs to be able to call the functions
> > type_natural_mode,
> > classify_argument, examine_argument and contains_aligned_value_p,
> > which are
> > defined in config/i386/i386.c. This patch makes them non-static (and adds
> > declarations - otherwise bootstrap breaks). I've been resisting
> > sending this
> > patch because (1) in the long run I believe I can rework the plugin to
> > not
> > need access to these functions; and (2) it is of no intrinsic interest
> > to gcc:
> > it only benefits my plugin. However it would clearly make life much
> > easier for
> > my users, since they would then not need to build their own patched
> > version of
> > gcc. OK to apply to the 4.5 branch and mainline?
> >
> > Ciao,
> >
> > Duncan.
> > <i386_static.diff>