Re: GDL2 - patches for OSX (2)

[Stéphane Corthésy]

Hi,

On Wednesday, Mar 26, 2003, at 09:45 Europe/Zurich, Manuel Guesdon 
wrote:

> Hi Stéphane,
>
> On Tue, 25 Mar 2003 23:00:42 +0100 Stéphane Corthésy 
> <stephane@sente.ch> wrote:
> > | Index: EOOrQualifier.m
> > | ===================================================================
> > | RCS file:
> > | /cvsroot/gnustep/gnustep/dev-libs/gdl2/EOControl/EOOrQualifier.m,v
> > | retrieving revision 1.7
> > | diff -r1.7 EOOrQualifier.m
> > | 56c56
> > | <   return [[[self alloc] initWithQualifierArray: array] 
> > autorelease];
> > | ---
> > |  >   return AUTORELEASE([[self alloc] initWithQualifierArray: 
> > array]);
> > | 89c89
> > | <   NSMutableArray *qualArray = [NSMutableArray array];
> > | ---
> > |  >   NSMutableArray *qualArray = [[NSMutableArray alloc] init];
>
>
> For this one, I usually prefer:
>         NSMutableArray *qualArray = (NSMutableArray*)[NSMutableArray 
> array];
> because:
>         o we don't need to know how NSMutableArray make it's 
> allocation and initialization
>         o if we want to track memory leaks, it's easier to make a grep 
> on alloc and check cases
> if there's not too much ones (i.e. if alloc are mainly done in objects 
> constructors).


Tracking memory leaks and zombies is not as easy as searching 
alloc/release/autorelease/etc. It's much more complicated than this. By 
creating some many autoreleased objects you're shooting yourself in the 
foot IMHO. You'll get in troubles when it comes to find zombies...
One of the big problems of EOF is the large number of transient objects 
created during operations: the more autoreleased object you create, the 
slower your application will become, and the harder it is to correct 
zombies.

While I don't agree with you on that point, I guess that other gnustep 
people think the way you do, so I'll stop coming again and again on 
that.

Stéphane