From: Claus Fischer <cfischer AT tcadcs6 DOT sc DOT intel DOT com>
Newsgroups: gnu.gcc.help,comp.os.os2.programmer.misc,comp.os.msdos.djgpp
Subject: Re: C++ constructor problem solved!! Read how, it might come in useful in your future programs!
Date: 20 Jan 1998 08:45:59 -0800
Organization: Intel Corporation
Lines: 24
Message-ID: <j5lafcr3xh4.fsf@tcadcs6.sc.intel.com>
References: <bWLoegW7sFse-pn2-bxyqvlUyVUyY AT localhost>
	<34C2817A DOT 5547BE86 AT alcyone DOT com>
	<bWLoegW7sFse-pn2-Y8HSrHrj4UXy AT portA32 DOT Generation DOT NET>
NNTP-Posting-Host: tcadcs6.sc.intel.com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk


Gili wrote:
> 	I found a second (and perhaps better) way of doing this. Now I use 
> "this->fstream.fstream(<parameters>".

Perhaps this should be questioned in a standards newsgroup to
completely clarify it; however, I want to guard you; even though this
might work for a current version of your C++-compiler, it's quite
likely not going to work for the final version.

No matter which syntax you use to refer to
  fstream::fstream(<parameters>)
as a constructor - and the above form is just a syntactical variation
of accessing the constructor function - the compiler will probably be
required by the standard to create a temporary object.

I think, therefore, that the only reliable form to use the constructor
explicitely would be one of the alternatives suggested by previous
posts: either using a pointer to an external allocated fstream object,
or using the special form of try syntax which has been suggested by a
person who is apparently more knowledgeable in C++ than me:-)

Claus Fischer
Not speaking for Intel.