Unexpected "is_equal" failure

[Andrei Lascu]

In the attached test case, `r_0` are `r_1` should be equivalent by construction, albeit `r_0` is comprised of multiple uses of DeMorgan laws, and some internal isl calls which should not affect the observable representation of the set, as far as I'm aware. The expected `is_equal` on line 66 fails, however. Also, the union on line 63 should be a no-op, but was left in as it corresponds to the operation on line 58.

As the test is rather convoluted, I unfortunately don't have the time right now to translate it to C format. I've also attached the C++ header file I've used to compile. As mentioned before, this is quite an older file. I've tried using a newer header from Polly [1], but I couldn't find a suitable `isl::space` constructor.

Andrei

[1] https://github.com/llvm/llvm-project/blob/main/polly/lib/External/isl/include/isl/cpp-checked.h

[Sven Verdoolaege]

On Thu, Jul 29, 2021 at 07:10:29AM -0700, Andrei Lascu wrote:
>
> In the attached test case, `r_0` are `r_1` should be equivalent by
> construction, albeit `r_0` is comprised of multiple uses of DeMorgan laws,
> and some internal isl calls which should not affect the observable
> representation of the set, as far as I'm aware. The expected `is_equal` on
> line 66 fails, however. Also, the union on line 63 should be a no-op, but
> was left in as it corresponds to the operation on line 58.
>
> As the test is rather convoluted, I unfortunately don't have the time right
> now to translate it to C format. I've also attached the C++ header file
> I've used to compile. As mentioned before, this is quite an older file.
> I've tried using a newer header from Polly [1], but I couldn't find a
> suitable `isl::space` constructor.

Start from unit and call add_unnamed_tuple and/or add_named_tuple.
Also, using a header from Polly is not going to make it any easier for me,
unless it's identical to the one generated by isl master.

In any case, thanks for the report.

skimo