On 07/28/2017 01:32 PM, Richard Biener wrote:
> On Fri, Jul 28, 2017 at 1:19 PM, Jakub Jelinek <ja...@redhat.com> wrote:
>> On Fri, Jul 28, 2017 at 01:11:51PM +0200, Martin Liška wrote:
>>> Hello.
>>>
>>> Following patch fixed race where we first need to properly build
>>> all gimple_build_debug_bind (arg, var, NULL) and then DECL_HAS_VALUE_EXPR_P
>>> can be registered.
>>>
>>> Patch can bootstrap on ppc64le-redhat-linux and survives regression tests.
>>>
>>> Ready to be installed?
>>> Martin
>>>
>>> gcc/ChangeLog:
>>>
>>> 2017-07-10 Martin Liska <mli...@suse.cz>
>>>
>>> PR sanitizer/81340
>>> * sanopt.c (sanitize_rewrite_addressable_params):
>>
>> Missing description of what you've changed. Also, it isn't immediately
>> clear why this is needed. For force_gimple* etc. I'd understand that
>> setting DECL_VALUE_EXPR prematurely will result in gimplification of that
>> into the DECL_VALUE_EXPR, but gimple_build_assign or gimple_build_debug_bind
>> shouldn't do that. Or does the former invoke match.pd and that cares about
>> it?
>
> I think target_for_debug_bind () gets confused.
Exactly:
tree
target_for_debug_bind (tree var)
{
...
if (DECL_HAS_VALUE_EXPR_P (var))
return target_for_debug_bind (DECL_VALUE_EXPR (var));
...
There's fixed changelog entry:
Set VALUE_EXPR after a debug stmt is generated (PR sanitizer/81340).
gcc/ChangeLog:
2017-07-10 Martin Liska <mli...@suse.cz>
PR sanitizer/81340
* sanopt.c (sanitize_rewrite_addressable_params):
Set VALUE_EXPR after gimple_build_debug_bind because
it calls target_for_debug_bind and would confuse the function.
gcc/testsuite/ChangeLog:
2017-07-10 Martin Liska <mli...@suse.cz>
PR sanitizer/81340
* g++.dg/asan/pr81340.C: New test.
Martin
>
>> Jakub
