Re: c++/2628

Re: c++/2628

[nathan]

The following reply was made to PR c++/2628; it has been noted by GNATS.

From: nathan@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, pere@hungry.com
Cc:  
Subject: Re: c++/2628
Date: 20 May 2001 16:53:02 -0000

 Synopsis: Class constructor called when operator new returns null.
 
 State-Changed-From-To: open->closed
 State-Changed-By: nathan
 State-Changed-When: Sun May 20 09:53:02 2001
 State-Changed-Why:
     You are mistaken about how operator new signals an error
     condition.
     
     ::operator new (size_t)
     throws an exception on out of memory, and NEVER returns
     NULL.
     ::operator new (size_t) throw ()
     returns NULL on out of memory.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2628&database=gcc

Re: c++/2628

[Petter Reinholdtsen]

The following reply was made to PR c++/2628; it has been noted by GNATS.

From: Petter Reinholdtsen <pere@hungry.com>
To: gcc-gnats@gcc.gnu.org
Cc: nathan@gcc.gnu.org
Subject: Re: c++/2628
Date: Sun, 20 May 2001 23:44:38 +0200

 [nathan@gcc.gnu.org]
 > You are mistaken about how operator new signals an error
 > condition.
 
 Nope, I knew all this.  I am just _very_ supprised to discover how
 the compiler assumes ::operator new (size_t) will throw exceptions
 when using -fno-exceptions.
 
 I believe this is a bug.
 
 I've learned about the flag -fcheck-new since I filed this bugreport,
 and believe it should be set by default when using -fno-exceptions.