Re: Ada The Best Language?

[lutz@iks-jena.de (Lutz Donnerhacke)]

* Bertrand Augereau wrote:
>> What do you expect by reading the source code line? What does really
>> happen? Why can this remain in production software for years?
>
>The semantics is that it is an unsigned like any other unsigned, except
>its capacity is different. And I would do capacity checking before this
>affectation, anyway.

Almost all people I showed this line instantly respond, that it stores the
number of remaining packets. A few seconds later some people came up with
the question how to store this in an IP packet because there is not field
for. So they noticed the error due their profund knowledge of the context.

Nobody was irritated by unmatching typed (even the compiler ...)

-------------------- t.c --------------------
#include <stdio.h>

int main(int argc, char**argv) {
   struct {
      unsigned int b : 1;
      unsigned int c : 1;
      unsigned int d : 12;
   } tt;
   tt.b = 4;
   tt.c = 1;
   tt.d = 105;
   return printf("%d, %d, %d.\n", tt.b, tt.c, tt.d);
}
-------------------- t.s --------------------
.LC0:
	.string	"%d, %d, %d.\n"
main:
	pushl %ebp
	movl %esp,%ebp
	andl $-2,%edx
	orl $2,%edx
	andl $-16381,%edx
	orl $420,%edx
	pushl $105
	movl %edx,%eax
	shrl $1,%eax
	andl $1,%eax
	pushl %eax
	movl %edx,%eax
	andl $1,%eax
	pushl %eax
	pushl $.LC0
	call printf
	movl %ebp,%esp
	popl %ebp
	ret
---------------- reorder t.s ----------------
	pushl $105        ; last argument on the stack (trivial)
	;
	andl $-2,%edx     ; edx is unknown value (clear the lowest bit)
	orl $2,%edx       ; set the second lowest bit
	andl $-16381,%edx ; mask F..FC003 => edx = ?.. ?000000 00000010
	orl $420,%edx     ; set some bits    edx = ?.. ?000001 10100110
	;
	movl %edx,%eax    ; edx = ?.. ?000001 10100110
	shrl $1,%eax      ; eax = ?....?00000 11010011
	andl $1,%eax      ; eax = 1
	pushl %eax        ; third argument on the stack
	;
	movl %edx,%eax    ; eax = ?...?000001 10100110
	andl $1,%eax      ; eax = 0
	pushl %eax        ; second argument on the stack
---------------- reorder t.s ----------------

Ok, the compiler is truly irritated.

>> 'guess'. No more questions.
>
>That's a bit of a nasty reaction. You don't have to be this arrogant.
>I myself don't use this type of code often, and I don't have any standard
>right there.

Sorry, it sound like 'I do not have a standard, I learned by trial and error.'

>The closest thing to a ref manual I got here, MSDN, says that for aligning
>on a int boundary, you have to specify a bitfield size of zero, so it must
>imply that other bitfields are packed, which sounds logical.

Definitly.