Re: [PATCH v3 7/9] compile: New 'compile print'

[Pedro Alves <palves@redhat.com>]

On 05/03/2015 03:05 PM, Jan Kratochvil wrote:
> On Wed, 29 Apr 2015 17:49:50 +0200, Pedro Alves wrote:
>> On 04/11/2015 08:44 PM, Jan Kratochvil wrote:
>>
>>> +    case COMPILE_I_PRINT_ADDRESS_SCOPE:
>>> +    case COMPILE_I_PRINT_VALUE_SCOPE:
>>> +      fputs_unfiltered ("#include <string.h>\n"
>>
>> OOC, why do we need the include?
> 
> Added:
> +      /* <string.h> is needed for a memcpy call below.  */
> 

Thanks.

>>>  
>>> +/* Fetch the type of COMPILE_I_EXPR_PTR_TYPE and COMPILE_I_EXPR_VAL
>>> +   symbols in OBJFILE so we can calculate how much memory to allocate
>>> +   for the out parameter.  This avoids needing a malloc in the generated
>>> +   code.  Throw an error if anything fails.
>>> +   Set *OUT_VALUE_TAKE_ADDRESSP depending whether inferior code should
>>> +   copy COMPILE_I_EXPR_VAL or its address - this depends on __auto_type
>>> +   array-to-pointer type conversion of COMPILE_I_EXPR_VAL, as detected
>>> +   by COMPILE_I_EXPR_PTR_TYPE preserving the array type.  */
>>
>> This comment seems a bit stale.  At least, I don't see an
>> OUT_VALUE_TAKE_ADDRESSP parameter.
> 
> OK, yes, updated.
> 
>    Function returns NULL only for COMPILE_I_PRINT_ADDRESS_SCOPE when
>    COMPILE_I_PRINT_VALUE_SCOPE should have been used instead.  

What does "should have been used instead" mean?  Is that a bug in the
caller?

>    This depends on __auto_type array-to-pointer type conversion of
>    COMPILE_I_EXPR_VAL, as detected by COMPILE_I_EXPR_PTR_TYPE preserving 
>    the array type.  */

Thanks,
Pedro Alves