On Wed, Mar 28, 2018 at 04:14:45PM +0200, Martin Liška wrote: > Hi. > > I'm sending Jakub's patch, where I removed the guard in asan_poison_variable. > Patch can bootstrap on ppc64le-redhat-linux and survives regression tests. > > Martin > > gcc/ChangeLog: > > 2018-03-28 Jakub Jelinek <ja...@redhat.com> > Martin Liska <mli...@suse.cz> > > PR sanitizer/85081 > * gimplify.c (asan_poison_variable): Don't do the check for > gimplify_omp_ctxp here. > (gimplify_decl_expr): Do it here. > (gimplify_target_expr): Likewise. > > gcc/testsuite/ChangeLog: > > 2018-03-28 Jakub Jelinek <ja...@redhat.com> > Martin Liska <mli...@suse.cz> > > PR sanitizer/85081 > * g++.dg/asan/pr85081.C: New test.
I guess it is ok for trunk for now, but for stage1 I'd like analysis on why you've added that !gimplify_omp_ctxp stuff at all. E.g. for local vars inside of OpenMP/OpenACC constructs there should be no reason why they couldn't be unpoisoned and poisoned inside of the OpenMP region. Jakub