On Thu, Oct 15 2020, Martin Liška wrote:
> Hello.
>
> As mentioned in the PR, since 74ca1c01d02e548f32aa26f9a887dcc0730703fb we
> wrongly
> merge a pair of VRP ranges with a different TREE_TYPE (in the PR one is char
> and second
> one is short). Then we merge 2 ranges and make an anti-range and bad things
> happen.
>
> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
>
> Ready to be installed?
> Thanks,
> Martin
>
> gcc/ChangeLog:
>
> PR ipa/97404
> * ipa-prop.c (struct ipa_vr_ggc_hash_traits):
> Compare types of VRP min and max as we can merge ranges of
> different types.
>
> gcc/testsuite/ChangeLog:
>
> PR ipa/97404
> * gcc.c-torture/execute/pr97404.c: New test.
> ---
> gcc/ipa-prop.c | 6 +++-
> gcc/testsuite/gcc.c-torture/execute/pr97404.c | 28 +++++++++++++++++++
> 2 files changed, 33 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/gcc.c-torture/execute/pr97404.c
>
> diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
> index 63c652f4ffc..3d0ebefa4aa 100644
> --- a/gcc/ipa-prop.c
> +++ b/gcc/ipa-prop.c
> @@ -123,7 +123,11 @@ struct ipa_vr_ggc_hash_traits : public ggc_cache_remove
> <value_range *>
> static bool
> equal (const value_range *a, const value_range *b)
> {
> - return a->equal_p (*b);
> + return (a->equal_p (*b)
> + && types_compatible_p (TREE_TYPE (a->min ()),
> + TREE_TYPE (b->min ()))
> + && types_compatible_p (TREE_TYPE (a->max ()),
> + TREE_TYPE (b->max ())));
I think the best check is:
types_compatible_p (a->type (), b->type ())
...which internally checks just the minima, but I understand the
intention is that the types are the same or at least compatible.
OK with that change.
Thanks!
Martin
