vlm for linux on x86_64
if you don't know what it is, you probably don't want it...
based on the admirable work of Brad Parker who changed the alpha assembler
code generator to produce c-code runnable on an x86_64 platform
this version uses a TUNTAP tap device for networking, that way
it doesn't need root privileges to run if you define a tap
interface beforehand. I decided to use a tap (as opposed to a tun)
interface because access to the complete ethernet frame is possible
with a tap interface - making my life easier since I'm currently
writing a chaosnet network module for Linux and would need access to
the complete ethernet frame therefore. Also it is possible to put tap
devices on a bridge with different mac addresses to be able to run
multiple vlm instances communicating to each other via chaosnet.
this version has fixes for:
- X11 with xcb lib (runs with or without xcb)
since clx directly accesses the bar-bones X11-protocol, xcb
gets confused when you try to XCloseDisplay/XSynch because clx tries
to keep track of the X11 calls internally which it is unable to
do with a mixture of XLIB calls and direct access to the X11-protocol.
This in turn keeps the vlm from cleanly closing a display, which it
uses when terminating and/or saving a LISP world.
So, if we detect xcb being used, we use a workaround for closing
X displays...
- keyboard mapping for the cold-load/debugger window
I deemed it unsatisfactory to not being able to use the CTRL-, META-,
SUPER-, and HYPER- keys in the cold-load/debugger window, so I did setup
a mapping for those keys like it is done in the GENERA system itself. You
may use the same keys now in the cold-load/debugger window.
- rational division (/ 3) yields 1/3 instead of 0
doesn't work in Brads version, because rational division ought to
produce a trap if there is a non-zero remainder from a division. I've
decide to produce asm code for rational division by generating a new
pseudo alpha instruction (x64ratquo) and generating asm code from that.
see alpha-assembler/imacmath.lisp and stub/process.lisp
- double float exceptions (/ 0.0d0) results in a valid trap now
has not been implemented in the alpha assembly version (forgotten
perhaps).
Changed that in the generation of alpha assembly code (tested on an alpha
as well). see alpha-assembler/imactrap.lisp.
there are extensions for:
- blocksize settable for attach-disk-channel (to be able to use fep-file
images with non-2048 word blocks, like images derived from ancient
ESDI disks.
- a unix crypt routine
links to unix crypt to enable use of modern crypt algorithms.
- unix cwd and unix home dir to be found in embedded communication area
You won't be able to use these extensions if you do not patch and
recompile respective the LISP definitions as well.
to install follow the instructions given in the file INSTALL.
./configure --help will give you some configuration options