Re: [PATCH 1/3] [gdb/exp] Fix target type of complex long double on arm

[Tom de Vries <tdevries@suse.de>]

On 5/29/24 13:44, Tom de Vries wrote:
> On 5/28/24 17:11, Tom Tromey wrote:
>>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
>>
>> Tom> For arm-linux, complex long double is 16 bytes, so the target 
>> type is assumed
>> Tom> to be 8 bytes, which is handled by the "case 64", which gets us 
>> double
>> Tom> instead of long double.
>>
>> Tom> Fix this by searching for "long" in the name_hint parameter, and 
>> using long
>> Tom> double instead.
>>
>> Can you file a compiler bug report about this and then put the link in a
>> comment?
>>
> 
> Done ( https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115272 ).
> 

>> Tom> +      /* If both "double" and "long double" are 8 bytes, choose 
>> "double"
>> Tom> +         for "complex double" and "long double" for "complex long
>> Tom> +         double".  */
>> Tom> +      if (builtin_type (gdbarch)->builtin_long_double->length () 
>> == 8
>> Tom> +          && name_hint != nullptr && *name_hint != '\0'
>>
>> I don't think the \0 part of this check is needed.
> 
> Ah, I see, it's the needle is empty string case that is special, and 
> name_hint is the haystack argument.

I've fixed this in a v2 ( 
https://sourceware.org/pipermail/gdb-patches/2024-June/209753.html ), 
and made the comment a bit shorter.

I've also added a dwarf assembly test-case.

Furthermore, I've updated the commit message to make it clear that the 
missing explicit target type is standard-conformant (I'm not sure 
whether I realized that already when I sent v1).

Thanks,
- Tom