[NTLUG:Discuss] Thinking ALPHA

[Steve Baker]

cbbrowne at godel.brownes.org wrote:

> I always found it something of a convenience that Pascal would always "puke"
> on that sort of thing; 'tis quite useful to *know* that you've not initialized
> a variable.

Well, C/C++ compilers *try* to tell you - but programmers can get pretty
devious about how they do or don't initialise things:

  int x ;
  if (  y ) x = 1 ;
  if ( !y ) x = 2 ;    /* Ugh! */
  z = x ;

...x is always initialised - but the compiler will probably still
complain
that it isn't.

  int x ;

  do_something ( & x ) ;
  z = x ;

...the compiler can't tell whether 'x' is initialised by 'do_something'
or not...it's a literal impossibility because the code for
'do_something'
may not even have been written at the time when the offending snippet
is compiled.

So, it's pretty hard for a compiler to figure it out with any certainty
and (IMHO) people *hate* warning messages about things that are
perfectly
good code more than they hate the compiler not warning about things that
are marginally wrong.

It's a hard call for the compiler writer.

It's not a distinction between Pascal and C/C++ - it's a distinction
between compilers.  The SGI IRIX compiler complains about an entirely
different set of problems than the GNU compiler - and those are
different
again from the problems that the M$VC compiler will pick up.  Even after
my Tux code had been compiled by all three of those, there were still
warning messages popping up out of the Apple C++ compiler on the Mac
and Borland C++ under Windoze.

To echo the words of our speaker at the meeting yesterday: Porting code
shows up errors...and that's because (in part) the more compilers look
at your code, the more things are going to get warned about.

-- 
Steve Baker                  http://web2.airmail.net/sjbaker1
sjbaker1 at airmail.net (home)  http://www.woodsoup.org/~sbaker
sjbaker at hti.com      (work)