This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [DRAGONEGG] Expose 4 functions in i386.c
On Wed, Apr 14, 2010 at 11:41 AM, Duncan Sands <baldrick@free.fr> wrote:
> Hi Richard,
>
>>> I completely agree that target hooks are the way to go. ?That's not the
>>> issue here. ?The question is whether it is nonetheless ok to apply this
>>> to the gcc-4.5 branch in order to make things easier for my users, even
>>> though it is technically the wrong thing to do in the long term.
>>
>> No, we don't want this kind of kludges - you can apply them locally.
>
> I can't apply it locally, since I'm not distributing gcc (and it's gcc that
> needs to be patched), I'm only distributing a plugin. ?With this patch,
> users
> would be able to use the plugin with their system supplied gcc-4.5, without
> it
> they need to build their own (patched) copy of gcc-4.5. ?That said, I can
> understand that you don't want to take this patch - the reason I sent it so
> late is that I also think it sucks. ?But I was swayed by the thought that it
> would be particularly convenient for my users.
On many systems they will have to anyway as plugins are an
optional feature that can be disabled. Note that while the patch
in question might not look bad it establishes precedent for
a bad style - we should (do) have proper interfaces to such things,
not start exporting random functions (especially not out of backends!).
Richard.
> Ciao,
>
> Duncan.
>