[PATCH] noclobber & noescape annotations for function arguments

Richard Guenther richard.guenther@gmail.com
Thu May 6 09:14:00 GMT 2010


On Fri, Apr 16, 2010 at 5:40 PM, Manuel López-Ibáñez
<lopezibanez@gmail.com> wrote:
> On 16 April 2010 17:37, Manuel López-Ibáñez <lopezibanez@gmail.com> wrote:
>> But I do not see the fnspec simplifying things. On the contrary, it is
>> the same inflation/complexity and in addition this complexity is
>
> In fact, the fnspec is more complex in terms of usability, because for
> valid combinations of attributes it forces the user to remember
> additional characters.

I'm not exactly sure how to go forward.  I want to keep the easy access
to the information.  Thus, what about making the new attribute
internal-only (naming it "fn spec") and as followup add individual
attributes which then would during their processing add/rewrite the
"fn spec" attribute (and not insert itself)?

That allows the Frontends (I'm thinking of the Fortran frontend
obviously) to annotate some of their runtime while we can
continue bikeshedding about names and semantics of
attributes we want to expose to our users.

Would that be ok?

Thanks,
Richard.

> Cheers,
>
> Manuel.
>



More information about the Gcc-patches mailing list