[LCP](no subject)

[Joachim Bauernberger]

First sorry or the broken thread by me not putting a subject on the last 
mail. Finger must have slipped on hte send button ...

On Monday 22 April 2002 11:38, Vincent Penquerc'h wrote:
> > Also, I have never come across a sigbus in my code before so 
> > I tried to
> > compile 
> > it under libc6 to see if it  would make a difference. The 
> > result on libc6 is
> > a segv. 
> 
> This puzzles me a bit, as SIGBUS is sent when two things happen
> at the same time:
> 1 - a memory value is fetched which size is not aligned with the
>     bus (eg, a 32 bit value not aligned on 32 bits)
> 2 - the CPU can't do two reads to rebuild the value
> 
> i386 can do step 2, so there should never be SIGBUSes on i386.

Didn't Netscape 4.x have big problems with SIGBUS when migrating it from 
libc5 to libc6 on Linux. I remember seeing it several times before.
 
> Have a look at your headers, maybe SIGBUS has a value which is
> also allocated to another signal, and this might tip you on
> what's happening ?

I looked into it, but headers on both systems seem to be in sync with 
what the values of SIGBUS and SIGSEGV are concerend. 
On libc5 signal.h includes <linux/signal.h> which defines SIGBUS as 7 
and SIGSEGV as 11.
On libc6 signal.h includes  <bits/signum.h> which does the same.


> 
> Sorry for the HTML, can't stop it :/

No Problem :-)

Thanks.
Joachim

> 
> -- 
> Vincent Penquerc'h 
> 

-- 
PGP Public Key:
http://www.mXscan.org/pgp/public.html