Jan> I am stuck with no working lisp on my laptop :-/ The 2003-12 snapshot
Jan> provided by Ole is nice, but I've had strange problems with 2003-12
Jan> ("NIL fell through ETYPECASE expression" errors when ASDF loads compiled
Jan> files).

I've seen the same problem, and I suspect PCL, not the memory layout -
I think I did cross-check that it also shows up on stock 2003-12 under
RedHat 7.x.  I've posted my tentative diagnostics to the CMUCL
implementation list:
<http://article.gmane.org/gmane.lisp.cmucl.devel/4976>

Jan> I tried to compile and cross-compile the 2003-11 snapshot changing the
Jan> DYNAMIC_0_SPACE_START offset to 0x58000000, but all I got is segfaults
Jan> in the load-world stage. Apparently there is more to it than just
Jan> changing this value.

Yes, you have to change TARGET-DYNAMIC-SPACE-START in
x86/compiler/x86/parms.lisp as well and make sure it reallay gets into
the image that's doing the compilation. At one point that didn't seem
to happen to me, so cross compilation seems like the safest bet.

Here is how to sort of bootstrap the dynamic space change from the
distribution even without access to a working platform:

1. Change x86-validate.h by hand and compile lisp using internals.h
   and friends from the dist.

2. The resulting binary will not segfault but complain about dynamic
   space lossage. Use this image to build world having changed
   parms.lisp as required.

3. Recompille lisp, rebuild world, then do genesis.

Ole




Reply via email to