Re: [vta, vta4.4, trunk?, 4.4?] avoid modifying locus of expr in place

[Alexandre Oliva <aoliva at redhat dot com>]

On Aug 20, 2009, Richard Guenther <richard.guenther@gmail.com> wrote:

> On Thu, Aug 20, 2009 at 9:00 PM, Alexandre Oliva<aoliva@redhat.com> wrote:
>> On Aug 20, 2009, Richard Guenther <richard.guenther@gmail.com> wrote:
>> 
>>> On Thu, Aug 20, 2009 at 10:51 AM, Alexandre Oliva<aoliva@redhat.com> wrote:
>>>> Does this sound like the right approach to fix the problem? ÂIs this
>>>> desirable for trunk or 4.4 branches? ÂIt doesn't fix any codegen
>>>> problems in there AFAIK, and there's only one non-expand use of this
>>>> function, which might get extraneous line number information elsewhere
>>>> because of this behavior.
>> 
>>> The correct fix is to rewrite expansion to not re-build generic.
>> 
>> I can see how this is somewhat related with the problem at hand, but I
>> don't see how some change at expansion time could make up for an earlier
>> call of gimple_assign_rhs_to_tree() in tree-affine.c, that would add an
>> extraneous location to a shared EXPR tree. ÂCould you please clue me in?

> tree-affine.c doesn't care about locuses at all.  That
> gimpleAssign_rhs_to_tree in the end shouldn't set a locus on the tree.

Aaah, all right, then.

> ATM it only does because of the use in expansion.

So should cfgexpand define a static wrapper that calls a
non-locus-setting version of gimple_assign_rhs_to_tree, and call that
instead, and then set the locus *after* unsharing the trees?

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist      Red Hat Brazil Compiler Engineer