[Bug c/108520] New: [C2x] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume)

[Bug c/108520] New: [C2x] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume)

[pexu@gcc-bugzilla.mail.kapsi.fi]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

            Bug ID: 108520
           Summary: [C2x] ICE in nonnull_arg_p, at tree.cc:14372 with -O1
                    and above (gnu::assume)
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pexu@gcc-bugzilla.mail.kapsi.fi
  Target Milestone: ---

Hi.

Consider the following testcase.  The important aspects are i) optimized, ii)
array of structures containing multiple elements, iii) assume.

Using 13.0.1 20230124 (experimental).

# cat assume.c
static
void f0() {  }

[[ gnu::nonnull(1) ]]
void f1(void *)
  {
    const struct { void (*f)(); } af[] = {
      { f0 } , { f0 } , { f0 }
    };

    for (__typeof__(af[0]) * it = af, * const end = af + sizeof(af) /
sizeof(af[0]); it != end; ++it)
      {
        [[ gnu::assume (it->f) ]];
        it->f();
      }
  }

# gcc -c -std=gnu2x -Wall -Wextra -O1 assume.c
during GIMPLE pass: ethread
assume.c: In function 'f1':
assume.c:16:3: internal compiler error: in nonnull_arg_p, at tree.cc:14372

---

(gdb) bt
#0  <address> in fancy_abort(char const*, int, char const*) ()
#1  <address> in nonnull_arg_p(tree_node const*) [clone .cold] ()
#2  <address> in get_range_global (r=..., name=<optimized out>)
    at <...>/src/gcc/gcc/value-query.cc:330
#3  <address> in global_range_query::range_of_expr (this=<optimized out>,
r=..., expr=<optimized out>,
    stmt=<optimized out>) at <...>/src/gcc/gcc/value-query.cc:406
#4  <address> in gimple_infer_range::check_assume_func (this=<optimized out>,
call=<optimized out>)
    at <...>/src/gcc/gcc/gimple-range-infer.cc:87
#5  <address> in gimple_infer_range::gimple_infer_range (this=<optimized out>,
s=<optimized out>)
    at <...>/src/gcc/gcc/gimple-range-infer.cc:166
#6  <address> in infer_range_manager::register_all_uses (this=<optimized out>,
name=<optimized out>)
    at <...>/src/gcc/gcc/gimple-range-infer.cc:374
#7  <address> in infer_range_manager::has_range_p (this=<optimized out>,
name=<optimized out>,
    bb=<optimized out>) at <...>/src/gcc/gcc/gimple-range-infer.cc:273
#8  <address> in infer_range_manager::maybe_adjust_range (this=<optimized out>,
r=..., name=<optimized out>,
    bb=<optimized out>) at <...>/src/gcc/gcc/gimple-range-infer.cc:290
#9  <address> in path_range_query::adjust_for_non_null_uses (this=<optimized
out>, bb=<optimized out>)
    at <...>/src/gcc/gcc/gimple-range-path.cc:502
#10 <address> in path_range_query::compute_ranges (this=<optimized out>,
dependencies=<optimized out>)
    at <...>/src/gcc/gcc/gimple-range-path.cc:623
#11 <address> in back_threader::find_taken_edge_cond (this=<optimized out>,
path=..., cond=<optimized out>)
    at <...>/src/gcc/gcc/bitmap.h:955
#12 <address> in back_threader::maybe_register_path (this=<optimized out>,
profit=...)
    at <...>/src/gcc/gcc/tree-ssa-threadbackward.cc:248
#13 <address> in back_threader::find_paths_to_names (this=<optimized out>,
bb=<optimized out>,
    interesting=<optimized out>, overall_paths=<optimized out>,
    at <...>/src/gcc/gcc/tree-ssa-threadbackward.cc:371
#14 <address> in back_threader::find_paths_to_names (this=<optimized out>,
bb=<optimized out>,
    interesting=<optimized out>, overall_paths=<optimized out>,
    at <...>/src/gcc/gcc/bitmap.h:955
#15 <address> in back_threader::maybe_thread_block (this=<optimized out>,
bb=<optimized out>)
    at <...>/src/gcc/gcc/bitmap.h:955
#16 <address> in back_threader::thread_blocks (this=<optimized out>)
    at <...>/src/gcc/gcc/tree-ssa-threadbackward.cc:975
#17 <address> in pass_thread_jumps_full::execute (this=<optimized out>,
fun=<optimized out>)
    at <...>/src/gcc/gcc/tree-ssa-threadbackward.cc:1105
#18 <address> in execute_one_pass (pass=<optimized out>)
    at <...>/src/gcc/gcc/passes.cc:2644
#19 <address> in execute_pass_list_1 (pass=<optimized out>)
    at <...>/src/gcc/gcc/passes.cc:2753
#20 <address> in execute_pass_list_1 (pass=<optimized out>)
    at <...>/src/gcc/gcc/passes.cc:2754
#21 <address> in execute_pass_list (fn=<optimized out>, pass=<optimized out>)
    at <...>/src/gcc/gcc/passes.cc:2764
#22 <address> in cgraph_node::expand (this=<optimized out>)
    at <...>/src/gcc/gcc/context.h:48
#23 <address> in symbol_table::compile (this=<optimized out>)
    at <...>/src/gcc/gcc/cgraphunit.cc:2015
#24 <address> in symbol_table::finalize_compilation_unit (this=<optimized out>)
    at <...>/src/gcc/gcc/cgraphunit.cc:2302
#25 <address> in compile_file () at <...>/src/gcc/gcc/toplev.cc:471
#26 <address> in toplev::main(int, char**) ()
#27 <address> in main (argc=<...>, argv=<...>) at <...>/src/gcc/gcc/main.cc:39

[Bug middle-end/108520] [13 Regression] [C2x] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume)

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
   Last reconfirmed|                            |2023-01-24
            Summary|[C2x] ICE in nonnull_arg_p, |[13 Regression] [C2x] ICE
                   |at tree.cc:14372 with -O1   |in nonnull_arg_p, at
                   |and above (gnu::assume)     |tree.cc:14372 with -O1 and
                   |                            |above (gnu::assume)
             Status|UNCONFIRMED                 |NEW
   Target Milestone|---                         |13.0
                 CC|                            |mpolacek at gcc dot gnu.org
          Component|c                           |middle-end
     Ever confirmed|0                           |1

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Confirmed.

[Bug tree-optimization/108520] [13 Regression] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume and gnu::nonnull)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |tree-optimization
            Summary|[13 Regression] [C2x] ICE   |[13 Regression] ICE in
                   |in nonnull_arg_p, at        |nonnull_arg_p, at
                   |tree.cc:14372 with -O1 and  |tree.cc:14372 with -O1 and
                   |above (gnu::assume)         |above (gnu::assume and
                   |                            |gnu::nonnull)

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed, simplified C/C++ testcase:
```
static void f0() {  }

typedef void (*ft)();
struct st { ft f; };
#define array_s 3
[[ gnu::nonnull(1) ]]
void f1(void *)
  {
    struct st af[array_s] = {
      { f0 } , { f0 } , { f0 }
    };

    for (struct st *i = af, *e = af+array_s; i != e; i++)
      {
        [[ gnu::assume (i->f) ]];
        i->f();
      }
  }
```
get_range_global is assuming (bad pun) that it can call nonnull_arg_p even for
non-current function parm_decls.

[Bug tree-optimization/108520] [13 Regression] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume and gnu::nonnull)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aldyh at gcc dot gnu.org,
                   |                            |amacleod at redhat dot com,
                   |                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So, unless nothing in the ranger uses cfun but passes around struct function
*fun
on which it should operate, I think gimple_infer_range::check_assume_func needs
to
temporarily set_cfun to fun and switch it back later.
Generally set_cfun is fairly expensive, but for the assumption functions one
would hope that at least in most cases they have the same optimization and
target nodes as the "caller".

[Bug tree-optimization/108520] [13 Regression] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume and gnu::nonnull)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Slightly cleaned up testcase:
static void foo () {}
struct S { void (*f) (); };

[[gnu::nonnull (1)]]
void
bar (void *x)
{
  struct S a[3] = { { foo }, { foo }, { foo } };
  for (struct S *i = a, *e = a + 3; i != e; i++)
    {
      [[assume (i->f)]];
      i->f ();
    }
}

[Bug tree-optimization/108520] [13 Regression] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume and gnu::nonnull)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andrew Macleod <amacleod@gcc.gnu.org>:

https://gcc.gnu.org/g:99f3ad2e5b117ee79a6dcf97288261e2fa32ab4c

commit r13-5806-g99f3ad2e5b117ee79a6dcf97288261e2fa32ab4c
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Mon Feb 6 13:07:01 2023 -0500

    Add function context for querying global ranges.

    When processing arguments for assume functions, call get_global_range
    directly and utilize a function context pointer to avoid any assumptions
    about using cfun.

            PR tree-optimization/108520
            gcc/
            * gimple-range-infer.cc (check_assume_func): Invoke
            gimple_range_global directly instead using global_range_query.
            * value-query.cc (get_range_global): Add function context and
            avoid calling nonnull_arg_p if not cfun.
            (gimple_range_global): Add function context pointer.
            * value-query.h (imple_range_global): Add function context.

            gcc/testsuite/
            * g++.dg/pr108520.C: New.

[Bug tree-optimization/108520] [13 Regression] ICE in nonnull_arg_p, at tree.cc:14372 with -O1 and above (gnu::assume and gnu::nonnull)

[amacleod at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108520

Andrew Macleod <amacleod at redhat dot com> changed:

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

--- Comment #6 from Andrew Macleod <amacleod at redhat dot com> ---
fixed.