RE: [microblaze-uclinux] PPPD problem with fork()

[sgomes@xxxxxxxxxxx]

Thanks Carsten,

It seems OK without the size-optimization.
Probably there is a problem in the toolchain?

Stephane

"Bartsch Carsten" <cbartsch@xxxxxxxxxx>
Envoyé par : owner-microblaze-uclinux@xxxxxxxxxxxxxx

03/03/2009 16:26

Veuillez répondre à
microblaze-uclinux@xxxxxxxxxxxxxx

A	<microblaze-uclinux@xxxxxxxxxxxxxx>
cc
Objet	RE: [microblaze-uclinux] PPPD problem with fork()

Switching off size-optimization
(remove -Os compiler option) helps my test-app, all addresses are now the
same in parent and child process.
 
Carsten
 
 
De: owner-microblaze-uclinux@xxxxxxxxxxxxxx
[mailto:owner-microblaze-uclinux@xxxxxxxxxxxxxx] En nombre de sgomes@xxxxxxxxxxx
Enviado el: martes, 03 de marzo de 2009 15:06
Para: microblaze-uclinux@xxxxxxxxxxxxxx
Asunto: [microblaze-uclinux] PPPD problem with fork()
 

Hi, 

I currently use the petalinux system with MMU support. I use the sources
files retrieved from petalinux-mmu-v0.10.tar.gz and not the latest sources
files. 
When I try to use pppd application to communicate with a modem, I have
a bug. 
I use pppd with the connect option which gives pppd a command to
execute. This command is executed using a fork() and execv().
The problem is that after the fork() call, in the created child, the pointers,
pointing on strings in the parent, point on a NULL pointer.

But if I compile pppd with #define __uCLinux__  which implies
to use a vfork() instead a fork(), pppd runs properly.

Has someone met the same problem in pppd or in other applications using
a fork()? 


Regards, 

Stephane