On Mon, 30 Jun 2014, Jeff Law wrote:
On 06/29/14 03:22, Marc Glisse wrote:
After looking at PR 61597, I updated the 2 conditions to:
+ if ((TREE_CODE (valbase) == VAR_DECL
+ && !is_global_var (valbase))
+ || TREE_CODE (valbase) == PARM_DECL)
a PARM_DECL is a local variable and returning its address is wrong,
isn't it?
Right. It can live in either a caller or callee allocated slot.
The "caller" case scares me a bit. Is it really wrong to return the
address in that case? The slot still exists after returning if the caller
is responsible for it.
When I first glanced at the patch my thought was "why is this in the path
isolation pass?"
A very pragmatic reason is that you and Richard asked me to after v1 of
the patch... It doesn't matter that much to me where this goes.
But I see you want to modify the returned value to be NULL.
I'll have to look at why you want to do that, but at least I understand why
it's utilizing the path isolation code.
Originally I mostly wanted to avoid warning several times. Now that the
warning is in a pass that runs only once, it isn't that necessary (it
remains necessary to do something in the front-end to avoid warning again
in the middle-end). It is still an optimization (it probably helps remove
dead code), though I could replace 0 with an undefined variable.
--
Marc Glisse
