[219] in Pthreads mailing list archive
Re: Solaris
daemon@ATHENA.MIT.EDU (Owen Rees)
Wed Nov 29 08:06:45 1995
To: pthreads@MIT.EDU
In-Reply-To: Message from proven@MIT.EDU of Sat, 18 Nov 1995 12:48:40
-0500.
<9511181748.AA12625@yaz-pistachio.MIT.EDU>
Date: Wed, 29 Nov 1995 11:43:29 +0000
From: Owen Rees <rtor@ansa.co.uk>
Christopher Provenzano <proven@MIT.EDU> writes:
>
> > Has anyone built the mit pthreads package on Solaris v2.4, or similar?
>
> I have. What's the problem you are seeing?
> CAP
I have just built 1_60 beta4_1 on Solaris 2.4 with gcc 2.7.0 and found
the following problems:
1) 'ar' core dumps when trying to build the library, replacing it with
the GNU binutils version of 'ar' fixes this problem.
2) tests/test_fork hangs like this:
3% tests/test_fork
parent process 4040
Using another session I discovered that the child is '<defunct>' (as
would be expected if the child is getting the right answer):
rtor 4040 3688 80 15:50:10 pts/12 0:12 tests/test_fork
rtor 4042 4040 0 0:00 <defunct>
3) tests/bench_pipe fails
4% tests/bench_pipe
Error: child read
Killing (-9) the process causes the whole session to disappear. As with
other systems, pipe has to have special syscall code. I adapted the
SunOS4.1.3 pipe syscall (paste it in, take leading underscores off the
name!), but since bench_pipe uses fork, the overall effect is to hang
rather than fall over. At least kill just kills the process leaving you
back in the shell.
Everything else seems to be working.
A litle bit more probing with gdb suggests that the parent process in a
fork is stuck inside the system call somehow - I don't know enough
about Solaris to go beyond this point.
Owen Rees
<rtor@ansa.co.uk>, <URL:http://www.ansa.co.uk/Staff/rtor.html>
Information about ANSA is at <URL:http://www.ansa.co.uk/>.
