Re: [PATCH v4] Fix sanitizer attribute infrastructure to use standard TREE_LIST format [PR113264]

["hanwei (K)" <hanwei62@h-partners.com>]

On Wed, Feb 25, 2026 at 12:28 AM hstk30 <hanwei62@huawei.com> wrote:
> On Mon, Feb 9, 2026 at 12:28 AM Andrew Pinski <andrew.pinski@oss.qualcomm.com> wrote:
> >
> > On Thu, Sep 4, 2025 at 10:43 AM Kees Cook <kees@kernel.org> wrote:
> > >
> > > The __attribute__((__copy__)) functionality was crashing when 
> > > copying sanitizer-related attributes because these attributes 
> > > violated the standard GCC attribute infrastructure by storing 
> > > INTEGER_CST values directly instead of wrapping them in TREE_LIST like all other attributes.
> > >
> > > Wrap sanitizer attributes INTEGER_CST values in TREE_LIST structures 
> > > to follow the same pattern as other attributes. This eliminates the
> > > copy_list() crashes when copying sanitizer attributes:
> > >
> > > test.c:4:1: internal compiler error: tree check: expected tree that contains ‘common’ structure, have ‘integer_cst’ in copy_list, at tree.cc:1427
> > >     4 | __attribute__((__copy__(__tanh)));
> > >       | ^~~~~~~~~~~~~
> > > 0x859d06 tree_contains_struct_check_failed(tree_node const*, tree_node_structure_enum, char const*, int, char const*)
> > >         ../../gcc/gcc/tree.cc:9126
> > > 0x860f78 contains_struct_check(tree_node*, tree_node_structure_enum, char const*, int, char const*)
> > >         ../../gcc/gcc/tree.h:3748
> > > 0x860f78 copy_list(tree_node*)
> > >         ../../gcc/gcc/tree.cc:1427
> > > 0xa755a5 handle_copy_attribute
> > >         ../../gcc/gcc/c-family/c-attribs.cc:3077
> >
> >
> > I am not a fan of the wrapping because it increases the memory usage 
> > slightly but it is required since the rest of the attributes code 
> > requires TREE_LIST here.
> >
> > So Ok.  I will do final testing either Monday or Tuesday and push it after that.
> 
> So this causes some ICEs in the testsuite:
> FAIL: c-c++-common/asan/inline-kernel.c   -O0  (internal compiler
> error: in tree_to_sanitize_code_type, at tree.cc:6704)
> 
> 0xa4376b fancy_abort(char const*, int, char const*)
>         /home/apinski/src/upstream-gcc-new/gcc/gcc/diagnostics/context.cc:1812
> 0x92f581 tree_to_sanitize_code_type(tree_node const*)
>         /home/apinski/src/upstream-gcc-new/gcc/gcc/tree.cc:6704
> 0x92f581 tree_to_sanitize_code_type(tree_node const*)
>         /home/apinski/src/upstream-gcc-new/gcc/gcc/tree.cc:6702
> 0x128bea2 print_no_sanitize_attr_value
>         /home/apinski/src/upstream-gcc-new/gcc/gcc/tree-cfg.cc:8223
> 0x128bea2 dump_function_to_file(tree_node*, _IO_FILE*, dump_flag)
>         /home/apinski/src/upstream-gcc-new/gcc/gcc/tree-cfg.cc:8276
> 
> 
> Looks like you forgot to update dump_function_to_file too.
> Can you double check all of the locations that use tree_to_sanitize_code_type to make sure they all have been fixed?

Fix this fail by below patch:

diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
index 66ea54f8b85..808b8000603 100644
--- a/gcc/tree-cfg.cc
+++ b/gcc/tree-cfg.cc
@@ -8220,7 +8220,10 @@ dump_default_def (FILE *file, tree def, int spc, dump_flags_t flags)
 static void
 print_no_sanitize_attr_value (FILE *file, tree value)
 {
-  sanitize_code_type flags = tree_to_sanitize_code_type (value);
+  /* Extract the INTEGER_CST from the TREE_LIST wrapper.  */
+  gcc_assert (TREE_CODE (value) == TREE_LIST);
+  sanitize_code_type flags = tree_to_sanitize_code_type (TREE_VALUE (value));
+
   bool first = true;
   for (int i = 0; sanitizer_opts[i].name != NULL; ++i)
     {


I have checked the patch which first introduces the tree_to_sanitize_code_type 
(https://inbox.sourceware.org/gcc-patches/20250813104306.20804-3-claudiu.zissulescu-ianculescu@oracle.com/)


> 
> Thanks,
> Andrew
> 
> >
> > Thanks,
> > Andrew