Resent-Date: Fri, 15 Mar 1996 18:05:44 -0500
Date: Fri, 15 Mar 1996 15:00:31 -0800 (PST)
From: Randy Chapman <chapman@cs.washington.edu>
To: java-linux@java.blackdown.org
Subject: libc.so.5.3.6
Message-Id: <Pine.OSF.3.91l.960315145031.21143I-100000@saul2.u.washington.edu>
Foo: foo-bar
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Resent-Message-Id: <"Gx3kQ3.0.2F7.pTVIn"@liquor>
Resent-From: java-linux@java.blackdown.org
Precedence: list
Resent-Sender: java-linux-request@java.blackdown.org


Just a word of warning:

I wouldn't expect libc.so.5.3.6 and java to get along well if any serios
multithreading, especially combined with io, is done.

Why?  libc.so.5.3.6 uses internal, non-thread-safe back doors into
the allocation routines, namely __libc_valloc.  As far as I can
tell, this cannot be overriden, either.

This should only show up if the lowlevel io funtions are active and
in memory allocation and a thread switch occurs and the new thread 
allocated something.

This is obviosuly a much more serious problem with kernel-supported 
threads, but luckilty java-linux doesnt use those yet.

I will be trying to work with HJL to resolve this problem.

--randy