[Openmcl-devel] Ways around the with-simple-restart bug?

Gary Byers gb at clozure.com
Tue Dec 7 05:20:18 UTC 2004



On Mon, 6 Dec 2004, John Wiseman wrote:

>
> On Dec 6, 2004, at 7:18 AM, Gary Byers wrote:
>
>  > Unfortunately, there hasn't been a new release of OpenMCL in an even
>  > longer time.  (It would take a few days to put together, and I
>  > haven't had a few days' time in many months.)
>
> It seems unlikely, but... Is there any way that any of us could help
> put a release together?
>
>
> John
>


I think that lots of things would be better if there was an "0.14.2-p2"
release; aside from the INVOKE-RESTART-INTERACTIVELY bug, there are
a dozen or so other things that have been "fixed in CVS" for several
months now.

There are a few other things (Helmut Eller got the FFIGEN stuff working
with GCC 3.x (3.3?), Rick Taube's GTK+ interfaces, other things that I'm
probably forgetting) that would be nice to package with a new release.
I think that just getting a patch release built (incorporating existing
bug fixes) is probably highest priority.

The couple of days that I think it would take basically involves
writing release notes, some amount of smoke-testing (sometimes not
enough smoke-testing) on Darwin and Linux, and packaging binaries.
(If that seems more like a few hours' work than a few days', I'd tend
to agree; somehow, every time I've alotted a few hours for it it's
wound up taking a few days.)

The way that Clozure (the machine) is set up requires a little bit of
security paranoia (Clozure the consulting company keeps some customer
data on clozure the machine, and "clozure.com" and
"openmcl.clozure.com" are the same physical box.)  There's a way to
set up SSH-based CVS write access that's a little funky but that
works pretty well, and it wouldn't take a long time to set up some
sort of reasonably secure FTP equivalent (knock  wood.)

In theory, the stuff that's in the main 0.14 CVS tree should just
need to have its version number changed and should be ready to go.
(I know that I've botched a few things: checked bug fixes into the
bleeding edge tree and forgot to check them into the main tree, and
I know that I've tried to un-botch those cases that I've caught ...
see why it takes a couple of days ?)  Theory and reality may be a
little out-of-synch.

If anyone's interested in doing this, that'd be great; please let
me know



More information about the Openmcl-devel mailing list