🚀 go-pugleaf

In article <PINKAS.92Sep2173635@skywalker.intel.com> pinkas@skywalker.intel.com (Israel Pinkas) writes:
>My understanding is that a routine is only allowed to have undocumented
>side effects if they will have no effect on the execution of the program.

Sounds fair to me.

>In my book, if evaluating 1+2 launches missles, it would be OK as long as
>(1) the missles did not destroy the currently executing computer, and (2)
>the lack of a missle in the launch bay did not cause the program to abort:)

(1) isn't allowed to affect any other output of the program either  :-)

>The question still boils down to this:
>Can the compiler consider the memory past the destination of a copy as
>scratch memory, or does strcpy GUARANTEE only to modify the memory that is
>necessary for copying the string?

How many times do you have to be told?  John Levine posted an example where
the proposed side effect would cause an illegal difference.

For another example, the expression x+y is supposed to add two variables
(if they have addable types) and is not supposed to zero out every other
variable in the program.

If you would like an interpretation ruling to say "Operations that are
defined by the standard shall not have other arbitrary side effects"
then write to the committee requesting one.  If your employer decides to
make the request as an official corporate concern, then it will probably
even get the ruling (I personally think) rather than being laughed off.

[Regarding someone's faulty example of a variable length structure hack]
>I don't know of a single compiler that guarantees to place two static
>variables which are declared in the same file next to each other.

Yeah, that example was faulty.  I think you know how to construct a good one.

>The variable length struct is a hack.  There is nothing in the language
>that prevents the compiler from performing bounds checking on accesses to
>name.

Uh, nothing except the standard's definitions of [], unary *, and binary +
operators.  :-R rhetoric.
Bounds checking is not allowed to affect output of a valid but ugly program.
--
Norman Diamond       diamond@jit081.enet.dec.com
If this were the company's opinion, I wouldn't be allowed to post it.
"Yeah -- bad wiring.  That was probably it.  Very bad."