Re: >>/<< 32 Bit wrong on 32 architechture

From: Benjamin Riefenstahl (Benjamin.Riefenstahl_at_epost.de)
Date: 12/30/03 

Date: Tue, 30 Dec 2003 13:30:31 +0100

Hi Gerhard,

> Benjamin Riefenstahl <Benjamin.Riefenstahl@epost.de> wrote in message
>> In C89 the result of
>>
>> a << b and a >> b
>>
>> is explicitly undefined if
>>
>> b >= (CHAR_BITS * sizeof(a)) || b < 0

g.hintermayer@inode.at (Gerhard Hintermayer) writes:
> ... but interesting, that
>
> printf("%d\n", 5<<32);
>
> does print zero (linux, gcc 3.2).

It could print 12345 and would still be correct according to the C
language rules. When the standard says "undefined", the compiler can
do anything. Or maybe I don't understand what you mean here.

> Regardless what C89 Standard says, this is quite an unusable result,

If you want some other result, you just need to write some other
expression, at least in C.

> So if compilers cannot handle this, should this be checked in
> tclExecute ?

Of course in Tcl, we can do whatever we want. Also Tcl is not quite
so speed-oriented when compared to C math expressions. So yes, we
could do that. Do you want to write a TIP?

benny

Quantcast