Re: [clisp-list] incompatibility between old clisp and new OS ?

[Hoehle, Joerg-C. <Joe...@t-...>]

Don Cohen wrote:
>I have a server running FC4 which recently rebooted and could no
>longer run my clisp server code.
>This server gets automatic FC4 updates, inc. the OS, but does not
>automatically reboot when it does.  The recent reboot was to
>2.6.16-1.2069_FC4
[...]
>In this case it's an old one, 2.31, but I don't have any
>reason to think a more recent one would behave any differently.

I believe you should try a newer clisp.  The reason is that when low-level changes in kernels affect the memory layout that applications get to see, CLISP is typically affected (e.g. randomize VA space, new addresses for shared libraries etc.).  Then CLISP's source code is modified to work under the new configuration.  That's why it may not work to run some old clisp in a new kernel.

This is just a guess.  Alas, I have no idea which (if any) these changes to CLISP are, in case you prefer to backport them to your 2.31, instead of using a newer version.

It's not nice at all to see a running program not survive an OS upgrade. I think that's a drawback of CLISP's typical 32bit address space memory partitioning.  If this matters much to you, and performance doesn't suffer, you may try to use the WIDE_SOFT version instead, for I believe this has no such dependencies.

>I suppose this is unrelated to the .mem, purely a function of the
>lisp.run.
Just start the lisp.run without .mem, and issue (ext:gc) or whatever at the prompt.

Regards,
	Jorg Hohle.