[Bug c/100993] New: ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[Bug c/100993] New: ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[cnsun at uwaterloo dot ca]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100993

            Bug ID: 100993
           Summary: ICE with -O2: Segmentation fault,
                    cgraph_update_edges_for_call_stmt(gimple*, tree_node*,
                    gimple*)
           Product: gcc
           Version: tree-ssa
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: cnsun at uwaterloo dot ca
  Target Milestone: ---

$ gcc-trunk -v
Using built-in specs.
COLLECT_GCC=gcc-trunk
COLLECT_LTO_WRAPPER=/scratch/software/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/configure
--enable-languages=c,c++,lto --enable-checking-yes --enable-multiarch
--prefix=/scratch/software/gcc-trunk --disable-bootstrap
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.0.0 20210609 (experimental) [master revision
:c23bc3c72:87f9ac937d6cfd81cbbe0a43518ba10781888d7c] (GCC)

$ cat mutant.c
__builtin_acc_on_device(dev) { return __builtin_acc_on_device(dev); }

$ gcc-trunk -O2 mutant.c
mutant.c:1:1: warning: return type defaults to ‘int’ [-Wimplicit-int]
    1 | __builtin_acc_on_device(dev) { return __builtin_acc_on_device(dev); }
      | ^~~~~~~~~~~~~~~~~~~~~~~
mutant.c: In function ‘__builtin_acc_on_device’:
mutant.c:1:1: warning: type of ‘dev’ defaults to ‘int’ [-Wimplicit-int]
during IPA pass: inline
In function ‘__builtin_acc_on_device’:
cc1: internal compiler error: Segmentation fault
0xf091b3 crash_signal
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/toplev.c:327
0xa826d0 cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraph.c:1729
0xf8bdcd fold_marked_statements
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/tree-inline.c:5384
0xf9bf87 tree_function_versioning(tree_node*, tree_node*, vec<ipa_replace_map*,
va_gc, vl_embed>*, ipa_param_adjustments*, bool, bitmap_head*,
basic_block_def*)
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/tree-inline.c:6457
0xcc1f1b save_inline_function_body
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/ipa-inline-transform.c:658
0xcc1f1b inline_transform(cgraph_node*)
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/ipa-inline-transform.c:750
0xe1c8f4 execute_one_ipa_transform_pass
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/passes.c:2290
0xe1c8f4 execute_all_ipa_transforms(bool)
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/passes.c:2337
0xa89639 cgraph_node::expand()
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraphunit.c:1821
0xa8aa5f expand_all_functions
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraphunit.c:1992
0xa8aa5f symbol_table::compile()
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraphunit.c:2356
0xa8d95b symbol_table::compile()
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraphunit.c:2269
0xa8d95b symbol_table::finalize_compilation_unit()
        /tmp/tmp.I0nMxdxnJz-gcc-builder/gcc/gcc/cgraphunit.c:2537
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

[Bug c/100993] ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100993

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jsm28 at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org
            Version|tree-ssa                    |12.0
           Keywords|                            |accepts-invalid,
                   |                            |ice-on-invalid-code

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I suppose we should reject definitions in the __builtin_* namespace?  Joseph,
can we?  It's fine to define 'memcpy', but it should not valid to define (or
even declare?) __builtin_memcpy.

[Bug c/100993] ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100993

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think at least in the past glibc used to add attributes to builtins through
typeof (__builtin_whatever) __builtin_whatever __attribute__((...));
so not sure if we can completely forbid it, but we certainly should forbid when
 it has incompatible type or what one defines the body for it.

[Bug c/100993] ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[joseph at codesourcery dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100993

--- Comment #3 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
Rejecting __builtin_* declarations was previously discussed in bug 32455.  
If we do, it would probably be worth testing building various software 
with the patched GCC to see what issues arise.

For many purposes, __sync_* and __atomic_* are built-in function 
namespaces just like __builtin_* (see builtins.c:is_builtin_name).  
However, functions with such names are definitely defined in libgcc and 
libatomic, and quite possibly might be defined elsewhere.

[Bug c/100993] ICE with -O2: Segmentation fault, cgraph_update_edges_for_call_stmt(gimple*, tree_node*, gimple*)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100993

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-03-23
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note a better testcase now is:
```
int __builtin_acc_on_device(int dev) { return __builtin_acc_on_device(dev); }

```