Re: [RFC v0 0/6] ASCII Command for output section

[Nick Clifton <nickc@redhat.com>]

Hi Ulf,

>>   What do you think of this alternative version of your patch ?

OK, I have now applied a tweaked version of the patch.

> A little closer look reveals that there is some oddities in the white space.
> Also, when the string in the ASCII command is longer than the specified
> length, only we need to insert a '\0' character at the end.
> 
> For this:
> 
>      len = size;
> 
> must be changed to
> 
>      len = size - 1;

Done.


> The size is specified by a "mustbe_exp"
> 
>      | ASCII '(' mustbe_exp ')' NAME
> 
>          {
>            /* 'value' is a memory leak, do we care?  */
>            etree_type *value = $3;
>            lang_add_string (value->value.value, $5);
>          }
> 
> I cannot judge the consequences of having a full expression, instead of an INT.
> 
> Will the "value->value.value" work, if we do
> 
>       ASCII (3 * 15) "Long string"

It works.  The area assigned will be 45 bytes.  I tested it, and included a version
of this expression in the new linker test.


> We do a malloc in the "etree_type *value = $3;" statement.
> This is thrown away without a "free".
> 
> Is that a problem anywhere?
> I guess most OSes would reclaim everything once the 'ld' terminates.
> There will not be that many ASCII statements in a linker command file.

Right.  As far as I am concerned a leak this small is not important.

Cheers
   Nick