[Openmcl-devel] Building on a G5

[Gary Byers]

On Thu, 7 Apr 2005, flip phillips wrote:

> I'm doing my damnedest to build the bleeding-edge openmcl on my G5
> under 10.3.x (I've had other issues building on one of our development
> machines under Tiger, anyone trying that too?:)
>

If anyone was using Tiger pre-releases, I hope that they wouldn't
discuss them publicly and violate their NDAs.  (Seriously.)

> Here's what I've done, here's what fails, etc.
>
> I've DL the latest non-bleeding openmcl-darwinppc-all-0.14.3.tar.gz
> and installed it in the appropriate place, set up appropriate
> environment variables, use the openmcl script
>  From this, I've verified that I can build a bootstrap image and full
> image.
>
> I want to play with the IDE, so I decide to play with the 14-dev branch
>
> I cvs checked it out.
> Got the interfaces at openmcl-darwin-interfaces-040505.tar.gz
> Copied over a dppccl.image from dppccl-image-050328.tar.gz

Good.

>
> 1) kernel builds just fine
> 2) I have to  (ccl:xload-level-0 :force) since, at least i -think- that
> the full cvs had some crufty FASL in it

There shouldn't be any FASLs in CVS, but it is necessary in this case
to recompile everything.

It's often a good idea to do something like:

shell> cd ccl
shell> find . -name '*.dfsl' | xargs rm

> 3) Now,  (ccl:compile-ccl) hurts me. I get:
>


> ;Compiling
> "/Applications/OpenMCL/ccl/compiler/PPC/PPC64/ppc64-arch.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc64-arch.dfsl"...
> ;Compiling "/Applications/OpenMCL/ccl/compiler/PPC/ppc-arch.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc-arch.dfsl"...
> ;Compiling "/Applications/OpenMCL/ccl/lib/ppcenv.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppcenv.dfsl"...
> ;Compiling "/Applications/OpenMCL/ccl/compiler/PPC/ppc-asm.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc-asm.dfsl"...
> ;Compiling
> "/Applications/OpenMCL/ccl/compiler/PPC/PPC32/ppc32-vinsns.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc32-vinsns.dfsl"...
> ;Compiling "/Applications/OpenMCL/ccl/compiler/PPC/ppc-lap.lisp"...
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc-lap.dfsl"...
> ;Compiling "/Applications/OpenMCL/ccl/compiler/PPC/ppc-backend.lisp"...
> ;Compiler warnings for
> "/Applications/OpenMCL/ccl/compiler/PPC/ppc-backend.lisp" :
> ;   Undefined function NEXT-NX-NUM-OPS, in an anonymous lambda form.
> ;   Undefined function NEXT-NX-DEFOPS, in an anonymous lambda form.
> ;Loading #P"/Applications/OpenMCL/ccl/bindarwin/ppc-backend.dfsl"...
>  > Error in process listener(1): Undefined function NEXT-NX-DEFOPS
> called with arguments () .
>  > While executing: #<Anonymous Function #x602679E>
>  > Type :GO to continue, :POP to abort.
>  > If continued: Retry applying NEXT-NX-DEFOPS to NIL.
> Type :? for other options.
>
> Any hints / clues here?
> -f

The missing macros (NX-NUM-OPS, NEXT-NX-DEFOPS) are defined in
"ccl:compiler;nxenv.lisp", which gets compiled to "ccl:bin;nxenv".
COMPILE-CCL may have seen a stale fasl file and decided not to compile
this, but it looks like:PPC-BACKEND should explicitly do an

(eval-when (:compile-toplevel) (require "NXENV"))

rather than just assuming that something else has already done so.

That's a bug, but the simplest workaround is to:

a) delete all old fasls
b) recompile everything (via (COMPILE-CCL T) and (XLOAD-LEVEL-0 :FORCE)

>

> --
> flip phillips, phd
> http://www.skidmore.edu/~flip/