Re: [Qemu-devel] [PATCH v4 04/12] fix qemu_alloc/qemu_free for linux-user subsystem

[Jean-Christophe Dubois]

Le samedi 20 juin 2009 09:01:04 vibi sreenivasan, vous avez écrit :
> >
> > This is true. It sounds that the various xxx-user subsystems are not
> > using the main qemu_malloc.c implementation.
> >
> > But this is just a fact ... I am not sure how I should interpret your
> > comment.
>
> since various xxx-user subsystems are not using the main qemu_malloc.c
> implementation , your changes would be having  a different effect than
> intended.

Well, I already asked a question about this issue on the list on June 8th (see 
attached email) but nobody answered my question.

Now the qemu_malloc function signature is just the same and the functionality 
seems quite equivalent. As I pointed out in the attached email the behavior is 
not the same (in particular, no abort() in case of error) but overall I don't 
really see a problem in using these qemu_malloc() instead of malloc() or 
calloc().

Now if you happen to have a clue on why the linux-user and bsd-user (and not 
darwin-user) targets need to have a different implementation I would be 
interested to know.

JC

>
> Thanks & Regards
> Vibi Sreenivasan
>
> > JC
> >
> > > Thanks & Regards
> > > Vibi Sreenivasan

> --- Begin Message ---
>
>
>
> Subject: 
>
> [Qemu-devel] question about qemu_malloc and friends in	xxx-user/mmap.c
>
>
>
>
> Date: 
>
> Mon, 8 Jun 2009 18:16:05 +0200
>
>
>
>
> User-agent: 
>
> KMail/1.11.2 (Linux/2.6.28-12-generic; KDE/4.2.2; x86_64; ; )
>
>
>
>
> Hi everybody,
>
> A little question about qemu_malloc(). There is one implementation that is 
> heavily debated on the list. This implementation is for the main qemu part.
>
> However there are also alternative implementation in xxx-user/mmap.c files for 
> example. These alternative implementations are not on par with the main 
> implementation when it comes to the behavior. In particular it does not 
> abort() on error for example. Is this perceived as a problem? Should all 
> qemu_malloc() implementation have the same behavior (whatever is finally agreed 
> upon on the list)?
>
> On the same kind of topic, these files do not offer some of the desired 
> function. In particular there are no implementation of qemu_strdup() or 
> qemu_strndud() and therefore the classical strdup() function is used in some 
> files. Is this other issue perceived as another problem to solve?
>
> Thanks
>
> JC
>
>
>
> --- End Message ---