Re: [PATCH 2/4] short circuit out of a few places where we would allocate zero bytes

From: H. Peter Anvin <hpa@zytor.com>
Date: Dec 28, 2005, 16:58 UTC
Message-ID: <43B2C44B.1070102@zytor.com>
In-Reply-To: <Pine.LNX.4.64.0512272036380.14098@g5.osdl.org>

Linus Torvalds wrote:

Show 14 quoted lines

> 
> On Tue, 27 Dec 2005, H. Peter Anvin wrote:
> 
>>Better yet, either always return NULL or allocate 1 byte in that case, to get
>>consistent behaviour.
> 
> Yes. However, if you do the "return NULL" case (which is nicest), you'll 
> have to wrap "free()" too. There are some libraries where passing "free()" 
> a NULL pointer causes a SIGSEGV.
> 
> That said, I think that would be preferable to changing the source code to 
> unnecessarily avoid zero-sized allocations. Having a "xfree()" to match 
> "xmalloc()" makes sense. 
>

Yeah, although that might break GNU code which uses xmalloc that is included (GNU doesn't have xfree.) The easiest is just to allocate 1 byte when the user asks for 0.

Anyone knows what GNU xmalloc does?

	-hpa