[uClibc]Strange long long division problem with gcc-3.2.x/uclibc on ARM

Christopher Hoover ch at murgatroid.com
Fri Mar 14 05:33:01 UTC 2003


> It works for me with on 2.4.19-rmk6

This was gcc 3.2.x?

> so my initial guess is either 2.5.59-rmk1 is busted, or perhaps
> the libgcc division code is using floating point and you didn't
> enable the kernel fpu emulator.  dunno.

Hmmm ... I definitely have FP emulation built into my kernel.

# date
Thu Jan  1 02:06:18 UTC 1970
# uptime
  2:06am  up  2:06, load average: 0.00, 0.00, 0.00
# grep -i float /var/log/messages
Jan  1 00:00:05 (none) user.warn klogd: Fast Floating Point Emulator
V0.9 (c) Peter Teichmann.


Does it not seem odd that the additional printf changes the behavior?

Cheers,
-ch



> -----Original Message-----
> From: Erik Andersen [mailto:andersen at codepoet.org] 
> Sent: Thursday, March 13, 2003 9:07 PM
> To: root
> Cc: uclibc at uclibc.org
> Subject: Re: [uClibc]Strange long long division problem with 
> gcc-3.2.x/uclibc on ARM
> 
> 
> On Thu Mar 13, 2003 at 07:56:50PM -0800, root wrote:
> > 
> > The appended little program crashes with gcc-3.2.2 and uclibc 0.9.19
> > under 2.5.59-rmk1.  (It also failed on an earlier 3.2 and uclibc
> > combo.)  
> 
> It works for me with on 2.4.19-rmk6
> 
>     $ uname -a
>     Linux winder 2.4.19-rmk6 #2 Sun Feb 2 12:28:26 MST 2003 
> armv4l unknown
>     $ /usr/arm-linux-uclibc/bin/arm-uclibc-gcc
>     -Wall -Os foo.c -o foo
>     $ ./foo 
>     50 / 10 = 5
> so my initial guess is either 2.5.59-rmk1 is busted, or perhaps
> the libgcc division code is using floating point and you didn't
> enable the kernel fpu emulator.  dunno.
> 
>  -Erik
> 
> --
> Erik B. Andersen             http://codepoet-consulting.com/
> --This message was written using 73% post-consumer electrons--
> 




More information about the uClibc mailing list