[PATCH] D41399: [CodeGen] Represent array members in new-format TBAA type descriptors

Hal Finkel via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Wed Dec 20 09:00:28 PST 2017


hfinkel added inline comments.


================
Comment at: test/CodeGen/tbaa-array.cpp:24
+// CHECK-DAG: [[TAG_A_i]] = !{[[TYPE_A:!.*]], [[TYPE_int:!.*]], i64 0, i64 4}
+// CHECK-DAG: [[TAG_C_i]] = !{[[TYPE_C:!.*]], [[TYPE_int:!.*]], i64 0, i64 16}
+// CHECK-DAG: [[TYPE_A]] = !{[[TYPE_char:!.*]], i64 4, !"_ZTS1A", [[TYPE_int]], i64 0, i64 4}
----------------
kosarev wrote:
> kosarev wrote:
> > hfinkel wrote:
> > > Shouldn't this access have a size of 4, and an access for c->x[2] have a size of 4 and a specific offset and c->x[j] have a size of 12 and an offset of zero? Why does this list a size of 16?
> > > 
> > > In any case, please add tests for:
> > > 
> > >   int *bar2(C *c) {
> > >     return c->x;
> > >   }
> > > 
> > >   int bar3(C *c) {
> > >     return c->x[2];
> > >   }
> > > 
> > >   int bar4(C *c, int j) {
> > >     return c->x[j];
> > >   }
> > > 
> > Indeed, the access size is wrong as we mistakenly inherit it from the base type. D41452 fixes this. Thanks for catching.
> Hal, in bar2() we don't really access memory. What do we want to check with it?
> Hal, in bar2() we don't really access memory. What do we want to check with it?

Ah, good point. Ignore that one.


Repository:
  rL LLVM

https://reviews.llvm.org/D41399





More information about the cfe-commits mailing list