freenode/#sicl - IRC Chatlog
Search
0:58:56
no-defun-allowed
What are the outputs for a TOP-LEVEL-ENTER-INSTRUCTION? I can't find it in the Cleavir book.
1:22:00
no-defun-allowed
One of the outputs is called and the output of that is AREFed to lookup a function in the global environment, so it appears different.
1:31:48
Bike
when the closure is created. for a top level enter that's probably when the fasl is loaded
7:07:08
no-defun-allowed
Okay, I think how I get how the static environment (with the HIR evaluator) works now. Just to check: if it's a top level enter, then the first 5 (sicl-compiler:+first-constant-index+) values in the static environment are installed by the host with some common functions, and the function cells corresponding to the function-names are installed after those. Otherwise, the initialize-closure instruction copies in the
7:07:08
no-defun-allowed
first 5 values from the caller, and takes the rest of the values in the static environment as inputs?
7:09:58
no-defun-allowed
Great, thanks. (common objects, rather, NIL and the code object are two of those.)
7:23:24
no-defun-allowed
Probably, I'm not sure if I know enough about the Java virtual machine to compile to it, and it's sort of something to pass time while I'm stumped with what to do on other projects.
8:00:47
no-defun-allowed
Indeed, but if you squint hard enough, it's more like a Smalltalk virtual machine than it isn't; except that the surrounding environment is awful and there's these weird things called "primitive types".
12:20:59
scymtym
i have the impression that ABCL does code walking without macro expansion to determine whether the method body contains CALL-NEXT-METHOD calls: https://github.com/armedbear/abcl/blob/bca11a225a64466c785e291d5d01f19fc8c09ac3/src/org/armedbear/lisp/clos.lisp#L2764 am i missing something?
12:22:37
scymtym
(and indeed, manually macroexpanding a method in McCLIM that "hides" a CALL-NEXT-CALL call in a macro fixes the problem i'm seeing)
12:27:22
Shinmera
I wrote a compiler for a proto-java targeting the jvm. Can't say I enjoyed the bytecode much.
12:30:10
scymtym
shka_: that may also be the case. i'm currently wondering whether there is a bug there. i'm asking because i can't see how this could work correctly
12:41:40
scymtym
i reported it as https://github.com/armedbear/abcl/issues/301 . i think i did my due diligence to ensure the problem is not on my end