This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [lto][patch] Strip more language dependencies from the IR
Hi,
On Mon, 25 Aug 2008, Diego Novillo wrote:
> > Yes. The point of my comment is that, unless the middle end
> > specifically distinguishes variably-modified vs. other types when
> > interpreting DECL_SIZE, keeping the size in other cases is redundant.
>
> Actually, we need to make the distinction between constant and
> non-constant DECL_SIZEs, so eliminating the constant cases would
> hamper the middle end. See, for instance tree-sra.c and
> tree-ssa-ccp.c.
>
> > Possibly, but it is part of the code to set the assembler names early,
> > taken from Michael Matz's patch.
>
> OK. Michael, do you think it would make sense to push this part of your
> patch to mainline before 4.4? If not, it's not a big deal.
It's part of the "robustification", i.e. that some c++ frontend routines
don't barf on things they never saw before. Without generating assembler
names early there's no purpose, and I'm not sure we want to do that for
4.4. In general I would say that robustness never hurts, but it's the
frontend maintainers call.
Ciao,
Michael.