[H-GEN] g++/egcs peculiarity

[Craig Eldershaw]

>>   int *first=new int(5);       // create an int
>>   int *second=new int[10](5);  // create 10 ints
>>  
>>   myclass *one=new myclass(1,2);      // create a myclass
>>   myclass *two=new myclass[10](1,2);  // create 10 myclasses

>Stroustrup, The C++ Programming Language, Section r.5.3.3, pg 499.
>...

Good point Andrae (note to self: rtfm ! :-)  

I actually checked the egcs www page, its FAQ and the last month or so
of bug reports - but not the text (*bonk*).  though admittedly there is
(last I heard) no formal ANSI standard as yet - just a draft.

>i.e. what you did with the int there wasn't correct C++ at the time the
>book was written.  

First reaction is why *not* allow it ?  Second thought is "Oh yeah,
Stroustrap hates C-style arrays".

I assumed that g++ passed it OK, so it must be right.  But I guess that
(a) the standard is still draft, and (b) gcc has allowed non-standard
things in the past.  

>Granted this reference was current May 1991 so there's plenty of time
>for it to change.

The 3rd edition (1997) of Stroustrup doesn't say one way or another
(they dropped the nice reference section).

>crash.cc:13: warning: initialization in array new

But is seems odd that (a) egcs (a later derivitive of g++) is less
flexible than its ancestor; (b) that it issues a warning rather
than an error in some cases; and (c) it actually works correctly (ie.
the initialisation tests out OK) for the 'int *second=new int[10](5);'
line.

Craig.


----------------------- HUMBUG General List --------------------------------
echo "unsubscribe general" | mail majordomo at humbug.org.au # To Unsubscribe