On Wed, Nov 09, 2016 at 02:16:45PM +0100, Martin Liška wrote:
> As shown in the attached test-case, the assert cannot always be true.
> Patch can bootstrap on ppc64le-redhat-linux and survives regression tests.
>
> Ready to be installed?
> Martin
> >From b55459461f3f7396a094be6801082715ddb4b30d Mon Sep 17 00:00:00 2001
> From: marxin <mli...@suse.cz>
> Date: Wed, 9 Nov 2016 11:52:00 +0100
> Subject: [PATCH] Remove unneeded gcc_assert in gimplifier (PR sanitizer/78270)
>
> gcc/ChangeLog:
>
> 2016-11-09 Martin Liska <mli...@suse.cz>
>
> PR sanitizer/78270
> * gimplify.c (gimplify_switch_expr):
No description on what you've changed.
That said, I'm not 100% sure it is the right fix.
As the testcase shows, for switch without GIMPLE_BIND wrapping the body
we can have variables that are in scope from the switch onwards.
I bet we could also have variables that go out of scope, say if in the
compound literal's initializer there is ({ ... }) that declares variables.
I doubt you can have a valid case/default label in those though, so
perhaps it would be simpler not to create live_switch_vars at all
if SWITCH_BODY is not a BIND_EXPR?
Jakub
