Viestiss� Lauantai 26. Hein�kuuta 2003 20:44, Gwenole Beauchesne kirjoitti:
> > and since the 0.6mdk gcc had no problem compiling the 5mdk kernel,
> > I would say it's the gcc update that wasn't good :-(
>
> I would say this is not a GCC bug.
>
> [EMAIL PROTECTED] vrac]$ rpm -q gcc
> gcc-3.3.1-0.7mdk
> [EMAIL PROTECTED] vrac]$ gcc -Wall -Wstrict-prototypes -Wno-trigraphs -O2
> -fno-strict-aliasing -fno-common -pipe -fomit-frame-pointer
> -mpreferred-stack-boundary=2 -march=i586 emufx.exp.c -c && echo $?
> 0
>
> (emufx.exp.c being the preprocessing output of the above-mentioned
> files with your options).

Sorry for the delay responding, but I had to try to rebuild with different 
options...

Now this is fun... or not ... :-(

To minimize the compile time I used:
LC_ALL=C rpm -ba kernel-2.4.spec --without smp --without secure --without 
enterprise --without doc --without source --without BOOT
(since it was on the 'up' kernel I got the "bug"...)

Try 1: - same options as when I got the bug above:
Result: fails with complaints about weird characters in module.c

Try 2: - change compiler options from -O2 to -Os
Result: compiles without problem...

Try 3: - change the -Os back to -O2
Result: compiles without problem...

Weird...
it seems to be a locale problem...
(I had to start using LC_ALL=C around the time of 2.4.21.1mdk..., 
never needed before...)

I noticed that I have a: /etc/rc.d/init.d/functions.rpmnew

wich has some changes to the locale / language settings, 
so I'll try to use the new settings and try to recompile a complete rpm set 

to see if the "bug" is still there, and if I still need "LC_ALL=C"

I'll let you know how it works out...


Regards

Thomas


Reply via email to