I’m not sure that machine-independent bytecode is going to be worthwhile, but I cleaned up and pushed the linklet-directory/bundle writing change. The part that Pycket needs to define is probably exactly what you have now, and then submodules should work.

We’ll take a look

One thing I notice is that the expander now uses fasl, but doesn’t expose the fasl functions

Can you add those as exports? Otherwise pycket will end up with two copies

And just to be clear, the new format will involve a fasl format with entries that are also in fasl format, right?

The expander doesn’t actually use racket/fasl; I just forgot to remove racket/fasl from require, and I’ll push that fix later. But if the expander eventually uses racket/fasl, I’d expect to just live with multiple copies.

@mflatt building the bytecode version of the expander crashes in the decompiler. The command is: ../../bin/racket bootstrap-run.rkt -c compiled/cache-src ++knot read read/api.rkt ++knot read read/primitive-parameter.rkt ++knot read read/readtable-parameter.rkt ++knot read read/readtable.rkt ++knot read read/number.rkt -s -x -B -o compiled/exp.zo and the error is Compiling and decompiling linklet to compiled/exp.zo application: not a procedure; expected a procedure that can be applied to arguments given: #f arguments...: '#hasheq((0 . #<linklet>)) context...: /home/samth/sw/plt/racket/src/expander/extract/decompile.rkt:8:0: compile-and-decompile5 "/home/samth/sw/plt/racket/src/expander/run.rkt": [running body] temp37_0 for-loop run-module-instance!125 for-loop [repeats 1 more time] run-module-instance!125 perform-require!78 Makefile:93: recipe for target 'bytecode' failed

I just pushed a fix

I guess that’s ok, since older versions of Racket won’t easily bootstrap anymore

Why did you add “cache-dir” to “.gitignore”? Is that created by something that I forgot about? (And if so, I wonder why it’s not in “compiled”.)

I think the alternative would be to add a primitive table for '#%linklet-expander since that doesn’t exist yet

I had cache-dir lying around, but it might be from before that went in compiled/ by default

I can remove that

@mflatt I wired everything up today, and it all works

now pycket -l racket/base uses everything from zo, which is very nice

(it does compile one trivial linklet whose content is (void) (void) but I don’t know where that comes from)