[Openmcl-devel] new OpenMCL snapshots available

[Gary Byers]

There are new "snapshot" tar archives for DarwinPPC[32|64] and for
LinuxX8664 available in the testing directry:

<ftp://clozure.com/pub/testing/openmcl-darwinppc-snapshot-060511.tar.gz>
<ftp://clozure.com/pub/testing/openmcl-linuxx8664-snapshot-060511.tar.gz>

[the "darwinppc" archive's name contained a typo until a few minutes ago.]

and there are 060511 archives containing heap images for LinuxPPC32 and for
LinuxPPC64 in that directory as well.

On all platforms: the FASL version changed, as did the version number
which links kernels and heap images.

The x86-64 version should be pretty nearly feature-complete.  Known
problems include:

  - FF-CALLs and callbacks which pass structures by value aren't
    handled correctly
  - "purification" of readonly code and static data still doesn't
    work (everything's still sloshing around in the dynamic heap.)

For SLIME users: the REPL seems to work, and sldb pops up a window
with a somewhat-correct backtrace in it.  Some other (possibly
obscure) things - I noticed some code that peeks inside a lexical
closure for the benefit of the SLIME inspector - will likely crash
on x86-64 (there needs to be a relatively stable support layer
to isolate SLIME/SWANK from this sort of stuff.)

For CLX users: the function XLIB::HOST-ADDRESS (in dep-openmcl.lisp
in telent CLX 0.7.3 and likely the older CLX version on clozure.com)
needs to be changed to call CCL:LOOKUP-HOSTNAME instead of 
CCL::HOST-AS-INET-HOST in order to avoid byte-ordering problems.

On DarwinPPC (haven't checked LinuxPPC), callbacks that receive more
than 13 floating-point args don't work correctly.

On all platforms, there are known problems with some backquote
constructs (details in bug-openmcl mail from several months ago.)

On all platforms, there are likely bugs that I'm either not thinking
of at the moment or not aware of.