This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: aggregate_value_p in the gimplifier
- From: Richard Henderson <rth at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: wilson at specifixinc dot com, gcc at gcc dot gnu dot org
- Date: Thu, 8 Jul 2004 10:04:57 -0700
- Subject: Re: aggregate_value_p in the gimplifier
- References: <10407081453.AA13459@vlsi1.ultra.nyu.edu>
On Thu, Jul 08, 2004 at 10:53:44AM -0400, Richard Kenner wrote:
> If we have MODIFY_EXPR whose RHS is a CALL_EXPR and they are of aggregate
> type, this really isn't an assignment since no copy will take place in the
> caller. Instead, we're just passing the address of the LHS to the call.
> So treating this as something we have to call memcpy for isn't quite
> right (and won't work because we'd try to make a variable-size temporary
> and either abort if checking is on or infinitely recurse if not).
I've been thinking about this recently, though I do not yet
have anything concrete to show for it. Please discuss any
proposed solution before committing anything.
r~