On Fri, Dec 16, 2016 at 03:16:14PM +0100, Marek Polacek wrote:
> On Fri, Dec 16, 2016 at 02:58:59PM +0100, Richard Biener wrote:
> > On Fri, Dec 16, 2016 at 2:03 PM, Bernd Schmidt <[email protected]> wrote:
> > > On 12/16/2016 12:49 PM, Marek Polacek wrote:
> > >
> > >> But as this testcase shows, this breaks when the default label shares a
> > >> label
> > >> with another case. On this testcase, when we reach the switch, we know
> > >> that
> > >> argc is either 1, 2, or 3. So by the time we reach vrp2, the IR will
> > >> have
> > >> been optimized to
> > >>
> > >> switch (argc) {
> > >> default:
> > >> case 3:
> > >> // argc will be considered 1 despite the case 3
> > >> break;
> > >> case 2:
> > >> ...
> > >> }
> > >
> > >
> > > Shouldn't we just remove the "case 3:" from the switch in this case? Would
> > > that fix things?
> >
> > We probably should indeed. But can we rely on this?
>
> I think we should do both -- apply my fix + investigated why we kept case 3
> around. I'm willing to look into this.
This is work of simplify_switch_using_ranges and then
11293 /* Update SWITCH_EXPR case label vector. */
11294 FOR_EACH_VEC_ELT (to_update_switch_stmts, i, su)
11295 {
11296 size_t j;
11297 size_t n = TREE_VEC_LENGTH (su->vec);
11298 tree label;
11299 gimple_switch_set_num_labels (su->stmt, n);
11300 for (j = 0; j < n; j++)
11301 gimple_switch_set_label (su->stmt, j, TREE_VEC_ELT (su->vec, j));
11302 /* As we may have replaced the default label with a regular one
11303 make sure to make it a real default label again. This ensures
11304 optimal expansion. */
11305 label = gimple_switch_label (su->stmt, 0);
11306 CASE_LOW (label) = NULL_TREE;
11307 CASE_HIGH (label) = NULL_TREE;
^^^^^^ this
But only after I wrote a patch I noticed that cfgcleanup merges default: and
case 3: before we reach .optimized, so I think there's nothing to do here.
Marek
