What's up with gcc 2.95.1 and "__va_arg_type_violation"

What's up with gcc 2.95.1 and "__va_arg_type_violation"

[Kevin Hendricks]

Hi,

When attempting to run the latest build of the jdk, I get the following error
message (but only on non-debug builds).

Unsatisfied Link error: __va_arg_type_violation

Figuring, I simply forgot some new piece of glibc 2.1.2, I used 

nm -o `find /lib -name "*"` | grep va_arg_type  
 
(for /lib, /usr/lib, /usr/local/lib, /opt/lib, etc)

and this damn routine simply does not exist on my machine.

Any ideas of where one might look to fulfill this thing.  This is new with
gcc-2.95.1 and/or glibc 2.1.2


By the way, native threads still do not work with the very latest glibc 2.1.2
because of the new semaphore functions.  If I force linking to the old
semaphore functions, everything starts working again.

I am building a debug glibc 2.1.2 and I will try to find the source of the
problem.

 Any ideas here would be greatly appreciated.

Thanks,

Kevin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: What's up with gcc 2.95.1 and "__va_arg_type_violation"

[schinder]

On Thu, Oct 07, 1999 at 09:08:16PM -0400, Kevin Hendricks wrote:
} 
} Hi,
} 
} When attempting to run the latest build of the jdk, I get the following error
} message (but only on non-debug builds).
} 
} Unsatisfied Link error: __va_arg_type_violation

If you turn off the optimizer, it will stop complaining.  I ran into this
building xemacs.

} 
} Figuring, I simply forgot some new piece of glibc 2.1.2, I used 
} 
} nm -o `find /lib -name "*"` | grep va_arg_type  
}  
} (for /lib, /usr/lib, /usr/local/lib, /opt/lib, etc)
} 
} and this damn routine simply does not exist on my machine.
} 
} Any ideas of where one might look to fulfill this thing.  This is new with
} gcc-2.95.1 and/or glibc 2.1.2
} 
} 
} By the way, native threads still do not work with the very latest glibc 2.1.2
} because of the new semaphore functions.  If I force linking to the old
} semaphore functions, everything starts working again.
} 
} I am building a debug glibc 2.1.2 and I will try to find the source of the
} problem.

Take a look at /usr/lib/ppc-redhat-linux/2.95.1/include/va-ppc.h.
It's in there.  I don't know why; I suspect it's part of an upcoming
fix of varargs.  The egcs that comes with Yellow Dog (and I assume
LinuxPPC) doesn't have this problem.

} 
}  Any ideas here would be greatly appreciated.
} 
} Thanks,
} 
} Kevin
} 

-- 
Paul Schinder
schinder@pobox.com

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: What's up with gcc 2.95.1 and "__va_arg_type_violation"

[Troy Benjegerdes]

Kevin Hendricks wrote:
> 
> Hi,
> 
> When attempting to run the latest build of the jdk, I get the following error
> message (but only on non-debug builds).
> 
> Unsatisfied Link error: __va_arg_type_violation
> 
> Figuring, I simply forgot some new piece of glibc 2.1.2, I used
> 
> nm -o `find /lib -name "*"` | grep va_arg_type
> 
> (for /lib, /usr/lib, /usr/local/lib, /opt/lib, etc)
> 
> and this damn routine simply does not exist on my machine.

Actually, this generally means you are attempting to pass a 'char' through
a va_args function. This is a "bad thing" since the char silently gets
converted to an integer, and the compiler can't type check it. I would like
for any character (or I think short int also) arguments to va_arg functions
and explicitly typecast them to a 32 bit value type in the function call.

Franz Sirl would be the guy to ask if you want to know the actually reasons
and full technical details for this.

-- 
--------------------------------------------------------------------------
| Troy Benjegerdes    |  troy@blacklablinux.com    |    hozer@drgw.net   |
|    Unix is user friendly... You just have to be friendly to it first.  |
| This message composed with 100% free software.    http://www.gnu.org   |
--------------------------------------------------------------------------

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/