[Bug tree-optimization/108698] New: [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[Bug tree-optimization/108698] New: [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

            Bug ID: 108698
           Summary: [13 Regression] decltype ((T() +
                    ‘excess_precision_expr’ not supported by
                    dump_expr<expression error>)) median(ndarray<T>) [with
                    T = double]’: since  r13-3290-g98e341130f87984a
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-linux-gnu

The following is isolated from python-scipy package:

$ cat scipy.ii
template <class> struct ndarray {};
template <class T> decltype(T() + 1.) median(ndarray<T>);
struct Trans_NS_functor_median {
  template <typename... Types> void operator()(Types...) { median(Types()...);
}
};
struct siegelslopes {
  void operator()(siegelslopes, double, void()) {
    ndarray<double> __trans_tmp_9;
    Trans_NS_functor_median{}(__trans_tmp_9);
  }
} siegelslopes0_y;
double siegelslopes0_x;
void siegelslopes0_method() {
  siegelslopes()(siegelslopes0_y, siegelslopes0_x, siegelslopes0_method);
}

$ g++ scipy.ii -c -m32 -std=c++14
scipy.ii: In instantiation of ‘decltype ((T() + ‘excess_precision_expr’ not
supported by dump_expr<expression error>)) median(ndarray<T>) [with T =
double]’:
scipy.ii:2:39: sorry, unimplemented: mangling excess_precision_expr
    2 | template <class T> decltype(T() + 1.) median(ndarray<T>);
      |                                       ^~~~~~

$ g++-12 scipy.ii -c -m32 -std=c++14

[Bug tree-optimization/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |13.0
     Ever confirmed|0                           |1
                 CC|                            |jakub at gcc dot gnu.org
           Priority|P3                          |P1
   Last reconfirmed|                            |2023-02-07
             Status|UNCONFIRMED                 |NEW

[Bug c++/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Simpler testcase:
template <typename T>
decltype (T () * T () + 1.0) foo ()
{
  return T () * T () + 1.0;
}

void
bar ()
{
  foo <float> ();
  foo <double> ();
  foo <long double> ();
}

The question is if the excess precision should be visible in the mangling in
some way or not.

[Bug c++/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 54423
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54423&action=edit
gcc13-pr108698.patch

Actually, if it is just EXCESS_PRECISION_EXPR around REAL_CST, that one can be
handled easily compatibly with what it used to do before.
I haven't been able to get EXCESS_PRECISION_EXPR with other operands so far.

[Bug c++/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I mean in mangling...

[Bug c++/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:b1ed0c9671b99c6b06cbb8c61be14cdec0998de8

commit r13-5751-gb1ed0c9671b99c6b06cbb8c61be14cdec0998de8
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Feb 9 09:31:55 2023 +0100

    c++: Mangle EXCESS_PRECISION_EXPR <REAL_CST> as fold_convert REAL_CST
[PR108698]

    For standard excess precision, like the C FE we parse floating
    point constants as EXCESS_PRECISION_EXPR of promoted REAL_CST
    rather than the nominal REAL_CST, and as the following testcase
    shows the constants might need mangling.

    The following patch mangles those as fold_convert of the REAL_CST
    to EXCESS_PRECISION_EXPR type, i.e. how they were mangled before.

    I'm not really sure EXCESS_PRECISION_EXPR can appear elsewhere
    in expressions that would need mangling, tried various testcases
    but haven't managed to come up with one.  If that is possible,
    we'd keep ICEing on it without/with this patch, and the big question
    is how to mangle those; they could be mangled as casts from the
    promoted type back to nominal, but then in the mangled expressions
    one could see the effects of excess precision.  Until we have
    a reproducer, that is just theoretical though.

    2023-02-09  Jakub Jelinek  <jakub@redhat.com>

            PR c++/108698
            * mangle.cc (write_expression, write_template_arg): Handle
            EXCESS_PRECISION_EXPR with REAL_CST operand as
            write_template_arg_literal on fold_convert of the REAL_CST
            to EXCESS_PRECISION_EXPR type.

            * g++.dg/cpp0x/pr108698.C: New test.

[Bug c++/108698] [13 Regression] decltype ((T() + ‘excess_precision_expr’ not supported by dump_expr<expression error>)) median(ndarray<T>) [with T = double]’: since r13-3290-g98e341130f87984a

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108698

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.