> On Oct 4, 2022, at 1:37 PM, Joseph Myers <jos...@codesourcery.com> wrote:
>
> On Tue, 4 Oct 2022, Qing Zhao via Gcc-patches wrote:
>
>> + { "strict_flex_array", 1, 1, false, false, false, false,
>> + handle_strict_flex_array_attribute, NULL },
>
> You're not requiring that the attribute be applied to a declaration here.
Okay, so, you mean I need to require the attribute to be applied to a
declaration as following:
{ "strict_flex_array", 1, 1, true, false, false, false,
handle_strict_flex_array_attribute, NULL },
?
Then when the attribute is attached to the type of the decl, the ICE will not
happen? (Yes, I tried, and the ICE was replaced with the following warning:
/home/opc/Install/lates/bin/gcc t.c
t.c:1:1: warning: ‘strict_flex_array’ attribute does not apply to types
[-Wattributes]
1 | int [[gnu::strict_flex_array(1)]] x;
| ^~~
Thanks, I will update the patch with this.
Let me know if you have more comments.
Qing
>
>> +static tree
>> +handle_strict_flex_array_attribute (tree *node, tree name,
>> + tree args, int ARG_UNUSED (flags),
>> + bool *no_add_attrs)
>> +{
>> + tree decl = *node;
>> + tree argval = TREE_VALUE (args);
>> +
>> + /* This attribute only applies to field decls of a structure. */
>> + if (TREE_CODE (decl) != FIELD_DECL)
>> + {
>> + error_at (DECL_SOURCE_LOCATION (decl),
>> + "%qE attribute may not be specified for %q+D", name, decl);
>
> But here you're using DECL_SOURCE_LOCATION on what might be a type, not a
> DECL. So if you have a test such as
>
> int [[gnu::strict_flex_array(1)]] x;
>
> that applies the attribute to a type, you get an ICE:
>
> t.c:1:1: internal compiler error: tree check: expected tree that contains
> 'decl minimal' structure, have 'integer_type' in
> handle_strict_flex_array_attribute, at c-family/c-attribs.cc:2526
> 1 | int [[gnu::strict_flex_array(1)]] x;
> | ^~~
>
> --
> Joseph S. Myers
> jos...@codesourcery.com
