With my 'fixed' version, eg 'mechdep_sys_lseek' moved to it's own file and
the 'sigprocmask/sigsuspend' removed from syscall.S it does work 99% of the
time.

I am getting C++ problems and a process termination problem so I am moving
to 1.4.1
to see if that does solve my problems.

Regards

Roger

----------
>From: Andy Finnell <andyf@vei.net>
>To: Roger Brown <rogerhb@xtra.co.nz>
>Cc: Christoph Badura <bad@oreilly.de>, port-mac68k@netbsd.org
>Subject: Re: libpthread.so.1.60 on NetBSD 1.3.3
>Date: Mon, Sep 6, 1999, 9:13 am
>

>Roger Brown wrote:
>> 
>> Still on 1.3.3...
>> 
>> When I remove sigprocmask and sigsuspend from 'config.flags' (where they are
>> mysteriously duplicated in the AVAILABLE_SYSCALLS clause but *that* does not
>> appear to be the problem)
>
>That's because I explicitly add them in the configure.in script.  Its
>needed for the 1.4.X systems so that sigprocmask and sigsuspend will be
>generated from the template file.
>
>> Then when the final .so is built, attempts to run programs comes up with
>> 
>> Undefined symbol 'SYS_compat_13_sigprocmask13' in libthread.so.1.60
>> 
>> Which shows that the functions that the assembler in syscall.S is trying to
>> call either don't exist or are not being generated.
>> 
>> Where is the 'compat' stuff supposed to come from?
>
>Look in /usr/include/sys/syscall.h for SYS_compat_13_sigprocmask13.  If
>its not in there look for all the different sigprocmask traps - one
>should be for 1.3.  After looking at the alpha port it looks like it
>should simply be SYS_sigprocmask.  You'll have to do the same for
>sigsuspend.  If that works, I seem to have a bug in my code, and I need
>to fix it.
>
> -andy