Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!unmvax!ncar!ames!think!barmar
From: barmar@think.COM (Barry Margolin)
Newsgroups: comp.windows.x
Subject: Re: ILA's X11 on Symbolics
Keywords: Xserver, Symbolics
Message-ID: <37713@think.UUCP>
Date: 21 Mar 89 20:46:50 GMT
References: <1738@trantor.harris-atd.com>
Sender: news@think.UUCP
Reply-To: barmar@brigit.think.com.UUCP (Barry Margolin)
Organization: Thinking Machines Corporation, Cambridge, MA
Lines: 42


We beta-tested ILA's X11, and I was very disappointed with the quality
of the version that they actually released, since most of the bugs I'd
reported hadn't been fixed.  Admittedly, during the beta test period
they were mostly concerned with the X-Remote-Screen (i.e. client)
side.

After they released a version for Symbolics to distribute I was able
to convince them to work on the more serious server bugs.  They've
fixed most of them, and an ECO is supposed to be distributed in the
near future.  The one serious bug that they haven't fixed is one that
prevents some pop-up menus (all the UWM menus, and the xterm
scroll-bar menu) from working.

The X11 server they distribute is simply a port of the MIT sample X
server from X11R2 (they're in the process of porting the R3 server).
The above bug is a bug that MIT knew about in the R2 server (I think
it has to do with byte swapping).  Most of the bugs that cause the
server to crash into the debugger are invalid C code that the
Symbolics environment catches: calling functions with the wrong number
of arguments (the compiler can't detect these when the call is through
a function pointer); accessing uninitialized structure components (the
code frequently passes structures that aren't fully initialized to
routines that try to swap bytes, and it dies when it tries to swap the
bytes of an uninitialized component).  One bug turned out to be a
Symbolics C compiler bug; it mis-compiles "foo = bar = baz", so foo
doesn't get the correct value; this caused deleted windows to stay on
the screen.

As I mentioned, most of the serious bugs are being fixed by the
patches that will be distributed soon.  We had been holding off paying
Symbolics for the product contingent on these bug fixes, and I sent
the order in last week.  It's not perfect, but it's usable now (I'm
using it at this very moment to read news and to compose this message
in GNU Emacs).


Barry Margolin
Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar