[Asterisk-Users] Build on Itanium fails

Steve Underwood steveu at coppice.org
Tue Aug 9 17:00:31 MST 2005


Jonas Arndt wrote:

> Dave,
>
> A segmentation fault is usually caused by the program writing in a 
> memory area that is not allocated (it could be a result of the 
> optimizer sometime as well). That means that it can potentially 
> overwrite code that are executing there. In worst case scenario you 
> could even cause a kernel Oops. A software that discovers that it is 
> missing a conf file will obviously have a potential of not working. 
> However, instead of writing in a memory area that it has not 
> allocated, it should inform about the situation in a log file

A kernel oops instead of a segfault? What kind of buggy OSes are you 
used to? Crazy memory access should only cause segfaults. The ability of 
an application to cause an oops would be considered a *very* serious OS bug.

Programs should not segfault when they have bad or missing config files. 
However, a large percentage do. If you haven't realised that you must 
have lived a sheltered life. :-)

> Now, if you are putting a software on the market that would take down 
> your system just because you are missing or having an incorrect 
> configuration file, you will find a lot of people seeing that as a 
> bug. Imagine ntp segfault and coredump just because the /etc/ntp.conf 
> is not there.

It can't bring down the system. Only the application itself.

> I think my problem is caused by the 32 versus 64 bits difference on 
> Itanium. Belive me, I have seen this before.

Do you realise that Asterisk runs on various 64 bit machines? x86_64, 
PPC, etc.

> Anybody out there with experience of compiling this animal on Itanium?
>
> By the way, I just built it and ran it on another architecture. 
> Without the conf files and there was no segmentation fault.

So what. This means nothing.

What not try a proper installation, and see what happens. Just to keep 
the rest of us idiots who haven't a clue about config files and proper 
operation happy.

Regards,
Steve





More information about the asterisk-users mailing list