|
|
 | | From: | toby | | Subject: | Re: RISC and CISC glass jaws | | Date: | 8 Jan 2005 13:13:54 -0800 |
|
|
 | Nick Maclaren wrote:
> ... it could be done without overhead for functions
> that did not refer to variables in outer functions
I think that much is indeed obvious. But taking the address of an inner
function that uses inner variables and invoking it outside of that
activation context is a runtime error that must surely be fatal.
Passing frame pointers makes no semantic difference; the frame simply
doesn't exist. But again, ignorance is my disclaimer.
--Toby
|
|
| | From: | Nick Maclaren | | Subject: | Re: RISC and CISC glass jaws | | Date: | 9 Jan 2005 10:39:39 GMT |
|
|
| In article <1105218834.562540.156930@c13g2000cwb.googlegroups.com>,
toby wrote:
>Nick Maclaren wrote:
>> ... it could be done without overhead for functions
>> that did not refer to variables in outer functions
>
>I think that much is indeed obvious. But taking the address of an inner
>function that uses inner variables and invoking it outside of that
>activation context is a runtime error that must surely be fatal.
>Passing frame pointers makes no semantic difference; the frame simply
>doesn't exist. But again, ignorance is my disclaimer.
Yes. Many languages that allow that prevent such use, statically or
otherwise - see the postings about Pascal.
Note that, even with solely parameters, there can be more than one
enclosing function frame corresponding to the same function.
Regards,
Nick Maclaren.
|
|
|
