On 2014-06-25 01:32, Marco Nembrini wrote:
Wouldn't an attribute like @nogc only be a keyword for attribute symbols, while something like nothrow is a keyword for everything?
I guess that's true.
E.g. using @nogc means I can't define my own "nogc" UDA but I can have a function or variable named nogc, while I can't have a function named "nogc". Being a "attribute keyword" seems a much smaller restriction on user code.
I guess so.
If what I wrote above is correct, why not declare existing compiler-attributes "attribute keywords", and then allow a mix of them: @(nothrow, public, const, "my_custom_attribute") void foo ();
Yeah, if the built-in attributes were implemented as UDA's, defined in object.d, then there would be less of a problem. Then one could always use fully qualified symbol names to disambiguate the attributes.
-- /Jacob Carlborg
