[Bug c/105150] New: [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[Bug c/105150] New: [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[gscfq@t-online.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

            Bug ID: 105150
           Summary: [9/10/11/12 Regression] ICE with -Ofast: verify_gimple
                    failed
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gscfq@t-online.de
  Target Milestone: ---

Affects versions down to r6, at -Ofast :


$ cat z1.c
#define A(name) __typeof (__builtin_##name (0)) name (); \
  float name##1 () { return !name (1); } \
  double name##2 () { return name (1.0L); }
#define B(name) A(name) A(name##l)
B (sqrt)


$ gcc-5 -c z1.c -Ofast
$
$ gcc-12-20220403 -c z1.c -Ofast
z1.c: In function 'sqrt1':
z1.c:2:36: warning: 'sqrt' argument 1 type is 'int' where 'double' is expected
in a call to built-in function declared without prototype
[-Wbuiltin-declaration-mismatch]
    2 |   float name##1 () { return !name (1); } \
      |                                    ^
z1.c:4:17: note: in expansion of macro 'A'
    4 | #define B(name) A(name) A(name##l)
      |                 ^
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c:5:4: note: built-in 'sqrt' declared here
    5 | B (sqrt)
      |    ^~~~
z1.c:1:49: note: in definition of macro 'A'
    1 | #define A(name) __typeof (__builtin_##name (0)) name (); \
      |                                                 ^~~~
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c: In function 'sqrt2':
z1.c:3:36: warning: 'sqrt' argument 1 type is 'long double' where 'double' is
expected in a call to built-in function declared without prototype
[-Wbuiltin-declaration-mismatch]
    3 |   double name##2 () { return name (1.0L); }
      |                                    ^~~~
z1.c:4:17: note: in expansion of macro 'A'
    4 | #define B(name) A(name) A(name##l)
      |                 ^
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c:5:4: note: built-in 'sqrt' declared here
    5 | B (sqrt)
      |    ^~~~
z1.c:1:49: note: in definition of macro 'A'
    1 | #define A(name) __typeof (__builtin_##name (0)) name (); \
      |                                                 ^~~~
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c: In function 'sqrtl1':
z1.c:2:36: warning: 'sqrtl' argument 1 type is 'int' where 'long double' is
expected in a call to built-in function declared without prototype
[-Wbuiltin-declaration-mismatch]
    2 |   float name##1 () { return !name (1); } \
      |                                    ^
z1.c:4:25: note: in expansion of macro 'A'
    4 | #define B(name) A(name) A(name##l)
      |                         ^
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c:5:4: note: built-in 'sqrtl' declared here
    5 | B (sqrt)
      |    ^~~~
z1.c:1:49: note: in definition of macro 'A'
    1 | #define A(name) __typeof (__builtin_##name (0)) name (); \
      |                                                 ^~~~
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
z1.c: In function 'sqrt1':
z1.c:5:4: error: mismatching comparison operand types
    5 | B (sqrt)
      |    ^~~~
z1.c:2:9: note: in definition of macro 'A'
    2 |   float name##1 () { return !name (1); } \
      |         ^~~~
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
int
double
if (1 == 0.0) goto <D.1991>; else goto <D.1992>;
z1.c:5:4: internal compiler error: 'verify_gimple' failed
    5 | B (sqrt)
      |    ^~~~
z1.c:2:9: note: in definition of macro 'A'
    2 |   float name##1 () { return !name (1); } \
      |         ^~~~
z1.c:5:1: note: in expansion of macro 'B'
    5 | B (sqrt)
      | ^
0xe5ecad verify_gimple_in_seq(gimple*)
        ../../gcc/tree-cfg.cc:5213
0xad1ac1 gimplify_body(tree_node*, bool)
        ../../gcc/gimplify.cc:16438
0xad1d67 gimplify_function_tree(tree_node*)
        ../../gcc/gimplify.cc:16509
0x8e2d87 cgraph_node::analyze()
        ../../gcc/cgraphunit.cc:675
0x8e6036 analyze_functions
        ../../gcc/cgraphunit.cc:1240
0x8e720d symbol_table::finalize_compilation_unit()
        ../../gcc/cgraphunit.cc:2500

[Bug tree-optimization/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2022-04-04
     Ever confirmed|0                           |1
                 CC|                            |jakub at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
   Target Milestone|---                         |9.5
           Priority|P3                          |P2
          Component|c                           |tree-optimization

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r6-4328-gc53233c660496efed5115ec1f30d3009ecf716d6

[Bug tree-optimization/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 52747
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52747&action=edit
gcc12-pr105150.patch

Untested fix.

[Bug tree-optimization/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #2)
> Created attachment 52747 [details]
> gcc12-pr105150.patch
> 
> Untested fix.

Hmm, we shouldn't detect those as sqrt and/or the frontend should not hand off
the wrongly typed arguments.  Not sure where the mismatch shows up though.

[Bug tree-optimization/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
So it's GENERIC folding and get_call_combined_fn as opposed to the GIMPLE
variant does not verify argument compatibility.  We have

 <call_expr 0x7ffff6524118
    type <real_type 0x7ffff653b348 double DF
        size <integer_cst 0x7ffff6517d50 constant 64>
        unit-size <integer_cst 0x7ffff6517d68 constant 8>
        align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff653b348 precision:64
        pointer_to_this <pointer_type 0x7ffff653b930>>
    readonly
    fn <addr_expr 0x7ffff66652c0
        type <pointer_type 0x7ffff6650dc8 type <function_type 0x7ffff6650d20>
            unsigned DI size <integer_cst 0x7ffff6517d50 64> unit-size
<integer_cst 0x7ffff6517d68 8>
            align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff6650dc8>
        readonly constant
        arg:0 <function_decl 0x7ffff65b1700 sqrt type <function_type
0x7ffff6650b28>
            readonly addressable used nothrow public external built-in decl_3
decl_5 decl_6 QI defer-output t.c:5:1
            align:8 warn_if_not_align:0 built-in: BUILT_IN_NORMAL:BUILT_IN_SQRT
            attributes <tree_list 0x7ffff65afb18
                purpose <identifier_node 0x7ffff653fbe0 nothrow>
                chain <tree_list 0x7ffff65afaf0
                    purpose <identifier_node 0x7ffff653fc30 leaf>
                    chain <tree_list 0x7ffff65afac8
                        purpose <identifier_node 0x7ffff65322d0 const tree_0
                            rid 0x7ffff65322d0 "const">>>>>
        t.c:5:1 start: t.c:5:1 finish: t.c:5:1>
    arg:0 <integer_cst 0x7ffff6537108 type <integer_type 0x7ffff65335e8 int>
constant 1>

where the CALL_EXPR_FN is an ADDR_EXPR of pointer-to unprototyped
and the decl is unprototyped as well but it got BUILT_IN_NORMAL:BUILT_IN_SQRT
and the inherited attributes.  IMHO that's a frontend problem - it should
not do that.

There's nothing get_call_combined_fn can do here if we do not want to
resort to check against the actual built-in decl that is registered
via builtin_decl_explicit ().

[Bug c/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|tree-optimization           |c

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
> IMHO that's a frontend problem - it should not do that.

Yes I think when we go and merge the builtin and the user supplied
(un-prototyped) decl, we should handle it similar to argument missmatch wtih
respect to not copying the builtins info.

[Bug c/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #6 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:71770a0ea920641c53912f725f5abd4413b38fd5

commit r12-8017-g71770a0ea920641c53912f725f5abd4413b38fd5
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Apr 6 10:07:26 2022 +0200

    gimple.cc: Adjust gimple_call_builtin_p and gimple_call_combined_fn
[PR105150]

    On Tue, Apr 05, 2022 at 11:28:53AM +0200, Richard Biener wrote:
    > > In GIMPLE, we call:
    > >   && gimple_builtin_call_types_compatible_p (stmt, fndecl)
    > > but that is insufficient, that verifies whether the arguments passed to
    > > GIMPLE_CALL match the fndecl argument types.  But that fndecl may very
well
    > > be the user declaration, so doesn't have to match exactly the builtin
    > > as predeclared by builtins.def etc. (though, there is the cotcha that
say
    > > for FILE * we just use void * etc.).  So e.g. in tree-ssa-strlen.cc
    > > we use additional:
    > >   tree callee = gimple_call_fndecl (stmt);
    > >   tree decl = builtin_decl_explicit (DECL_FUNCTION_CODE (callee));
    > >   if (decl
    > >       && decl != callee
    > >       && !gimple_builtin_call_types_compatible_p (stmt, decl))
    > >     return false;
    >
    > Yeah, I think we should use that (and only that) function decl
    > in get_call_combined_fn and gimple_call_combined_fn until the
    > frontend stops slapping wrong BUILT_IN_* on random decls.

    So, as a preparation step, this patch adjusts gimple_call_builtin_p
    and gimple_call_combined_fn so that they check argument types against
    the builtin_decl_explicit TYPE_ARG_TYPES rather than against the
    actual used fndecl.

    2022-04-06  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/105150
            * gimple.cc (gimple_call_builtin_p, gimple_call_combined_fn):
            For BUILT_IN_NORMAL calls, call
gimple_builtin_call_types_compatible_p
            preferrably on builtin_decl_explicit decl rather than fndecl.
            * tree-ssa-strlen.cc (valid_builtin_call): Don't call
            gimple_builtin_call_types_compatible_p here.

[Bug c/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #7 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:5df29fe79df659617793f955a1ea6c23a0617fe2

commit r12-8022-g5df29fe79df659617793f955a1ea6c23a0617fe2
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed Apr 6 16:47:47 2022 +0200

    gimple.cc: Follow-up to adjust gimple_call_builtin_p and
gimple_call_combined_fn [PR105150]

    On Wed, Apr 06, 2022 at 09:41:44AM +0100, Richard Sandiford wrote:
    > But it seems like the magic incantation to detect ârealâ built-in
    > function calls is getting longer and longer.  Can we not abstract this
    > in a single place rather than have to repeat the same long sequence in
    > multiple places?

    I've already committed it, so it can be only dealt with an incremental
    patch.
    Here is a patch that adjusts instead
gimple_builtin_call_types_compatible_p,
    after the assert:
      if (DECL_BUILT_IN_CLASS (fndecl) == BUILT_IN_NORMAL)
        if (tree decl = builtin_decl_explicit (DECL_FUNCTION_CODE (fndecl)))
          fndecl = decl;
    but we then lose the theoretical possibility of comparing against the
    actual user declaration.  Though I guess in the
    gimple-fold.cc
    gimple-low.cc
    gimple-match-head.cc
    calls to that function we also want this rather than what they do
currently.

    2022-04-06  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/105150
            * gimple.cc (gimple_builtin_call_types_compatible_p): Use
            builtin_decl_explicit here...
            (gimple_call_builtin_p, gimple_call_combined_fn): ... rather than
            here.

[Bug c/105150] [9/10/11/12 Regression] ICE with -Ofast: verify_gimple failed

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #8 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:973a2ce71f8dab559fbbfc34b59e39e047df74a6

commit r12-8041-g973a2ce71f8dab559fbbfc34b59e39e047df74a6
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Apr 7 09:10:58 2022 +0200

    tree.cc: Add tree_builtin_call_types_compatible_p [PR105150]

    And here is the follow-up patch that does the argument checking
    on GENERIC.  It ensures TYPE_MAIN_VARIANT == TYPE_MAIN_VARIANT
    compatibility on the arguments, except for pointer arguments
    where both builtin's prototype and actual arguments have to be
    pointers and satisfy tree_nop_conversion_p, and for promoted
    char/short arguments where argument need to have integral
    signed type tree_nop_conversion_p compatible with integer_type_node.

    2022-04-07  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/105150
            * tree.cc (tree_builtin_call_types_compatible_p): New function.
            (get_call_combined_fn): Use it.

            * gcc.dg/pr105150.c: New test.

[Bug c/105150] [9/10/11 Regression] ICE with -Ofast: verify_gimple failed

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[9/10/11/12 Regression] ICE |[9/10/11 Regression] ICE
                   |with -Ofast: verify_gimple  |with -Ofast: verify_gimple
                   |failed                      |failed

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk so far.
I'd strongly prefer not to backport though.

[Bug c/105150] [9/10/11 Regression] ICE with -Ofast: verify_gimple failed

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #10 from rguenther at suse dot de <rguenther at suse dot de> ---
On Thu, 7 Apr 2022, jakub at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150
> 
> Jakub Jelinek <jakub at gcc dot gnu.org> changed:
> 
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>             Summary|[9/10/11/12 Regression] ICE |[9/10/11 Regression] ICE
>                    |with -Ofast: verify_gimple  |with -Ofast: verify_gimple
>                    |failed                      |failed
> 
> --- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
> Fixed on the trunk so far.
> I'd strongly prefer not to backport though.

Agreed unless it shows up in real-world code.

[Bug c/105150] [10/11 Regression] ICE with -Ofast: verify_gimple failed

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.5                         |10.4

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 9 branch is being closed

[Bug c/105150] [10/11 Regression] ICE with -Ofast: verify_gimple failed

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.4                        |10.5

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 10.4 is being released, retargeting bugs to GCC 10.5.

[Bug c/105150] [11 Regression] ICE with -Ofast: verify_gimple failed

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.5                        |11.5

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10 branch is being closed.

[Bug c/105150] [11 Regression] ICE with -Ofast: verify_gimple failed

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105150

--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 87886 has been marked as a duplicate of this bug. ***