libera/#commonlisp - IRC Chatlog

8:38:42 chomwitt Hi. I hope my question is not irrelevant. I try StumpWM and i noticed that debian package has the lisp files included and so a package for ALL architectures. But arch has only a amd64 package with a binary and none lisp files.

8:39:01 chomwitt So how can in arch the sblc run the binary ?

8:39:14 chomwitt s/sblc/sbcl

8:42:14 beach If the binary is an executable, I suspect it won't need an installed SBCL.

8:45:33 chomwitt beach, yes. i checked it. its an ELF executable.

8:45:36 Nilby chomwitt: sbcl can save an exectutable that contains itself and all the loaded/compiled code that acts as relatively normal program.

8:46:07 chomwitt https://archlinux.org/packages/extra/x86_64/stumpwm/

8:47:07 beach chomwitt: I am not sure what you perceive as strange here.

8:48:25 chomwitt So SBLC is just a compiler ? There is not need of it to run stumpwm ?

8:48:59 beach chomwitt: There is an SBCL included in the executable as Nilby said.

8:49:34 chomwitt but why arch has sbcl as depedency ?

8:49:47 beach ACTION wouldn't know.

8:50:13 chomwitt beach, could it be that they dont know either ? :-)

8:50:32 beach Possibly.

8:54:23 chomwitt i havent test it in arch. In my devuan system i noticed that when running stumpwm and seeing it's version it reported 'ver 1.0.1 compiled in 5 Feb 2024'. That issue prompted my little quest because the version in the package was supposedly 22.11-3

8:54:44 chomwitt and in debian the executable is a script with : sbcl --load "/usr/lib/stumpwm/load.lisp"

8:55:19 chomwitt so being a clisp newbie that command seems to me like feeding code to an intepreter..

8:55:39 beach Er, clisp is not Common Lisp. It is an implementation of Common Lisp.

8:55:54 chomwitt sorry for that

8:56:06 beach Yes, it sounds like Debian is compiling from source and using an installed SBCL, whereas arch is distributing a binary containing everything.

8:56:35 beach Furthermore, SBCL does not use an interpreter by default. Everything is compiled.

8:57:59 chomwitt ok. thank you!

8:58:00 beach But it is likely that the Debian version is compiling some things at startup. Possibly just the load.lisp file.

8:58:38 beach That's not a problem though. The compiler is pretty fast.

8:59:44 chomwitt i see.

9:00:45 Nilby beach: I'm sorry to be a "well actually" guy, but sbcl does use an interpreter for throw away forms since it's faster.

9:01:25 beach Oh, OK. How is a "throw-away form" defined?

9:01:49 Nilby things that probably don't end up as code in a fasl

9:02:09 beach I need to think about that. Thanks for the information.

9:03:23 beach Nilby: If I do (defun f (x) x) at the REPL, and then (compiled-function-p #'f) I get T.

9:03:40 beach And I don't see any FASL involved here.

9:05:28 Nilby defun isn't thrown away

9:06:45 Nilby but if you just have like (+ 2 3) in file, it doesn't need to save it, and it's quicker not to compile it

9:08:57 beach I see.

9:09:37 Nilby i've read through it, but i don't remember it very well. it's called sb-fasteval, i think written by doug katzman a while back.

9:09:49 beach I believe you.

9:10:29 Nilby maybe it's only used when building sbcl itself, since i don't see sb-fasteval in my *features*

9:12:41 Nilby sbcl/src/interpreter/README talks about it's reasoning

9:18:51 Nilby i guess it's designed to speed up macro expansion but maybe not turned on in the build anymore

9:19:32 beach I wonder whether it is worth the trouble.

9:41:54 beach` ** NICK beach

10:45:25 antoszka_ ** NICK antoszka

10:45:25 alanz_ ** NICK alanz

10:45:25 gendl____ ** NICK gendl__

11:28:25 ski_ ** NICK ski

11:44:14 _death may be worth the trouble if you have something calling EVAL repeatedly with simple forms

11:58:47 dnhester26 I would like to add an option to a slot, does anyone know how to do this? E.g. like :initargs, I want something called :has-many that can accept a T value

11:58:50 dnhester26 http://metamodular.com/CLOS-MOP/initialization-of-slot-definition-metaobjects.html

11:59:18 dnhester26 That's the MOP doc that probably referes to what I'm talking about, but from there I don't really know what to do to do it

12:08:07 dnhester26 I started adding the MOP docs to the Technical Reference so that we can add examples as well given how hard it is to find them https://lisp-docs.github.io/cl-language-reference/meta-object-protocol/chap-6/6-1-sections#615-initialization-of-slot-definition-metaobjects

12:14:32 _death dnhester26: here's an old snippet that allows having a :history-max-depth option in slot specifiers https://gist.github.com/death/c4996d3a39559fb69d0f2c92f56e2bb5

12:14:56 _death dnhester26: another approach would be to avoid MOP and have your own defclass-like macro

12:16:14 _death I believe AMOP has more examples but don't have access to it at the moment

12:17:01 dnhester26 _death: thanks, I'll look into it

12:17:17 dnhester26 AMOP is a system like closer-mop?

12:19:25 _death AMOP is the book "The Art of the Metaobject Protocol"

12:24:01 dnhester26 ah, thanks, I just saw somewhere in chapter 3 they explain it, but only chapters 5 and 6 are available easily online

12:26:02 hexology- ** NICK hexology

12:40:03 Shinmera New release: https://shinmera.github.io/cl-gog-galaxy/ Support my continued work on Patreon: https://patreon.com/shinmera

12:47:14 dnhester26 _beach: thanks I can now create the slot, how can I then access it from the class? I can do inspect on (inspect (find-class 'foo)) and see it's there, but how can I access it from the code?

12:49:45 dnhester26 take it back

12:49:56 dnhester26 (has-foo-slot-p (car (c2mop:class-direct-slots (find-class 'foo))))

12:50:19 dnhester26 got it, I define the accessor in the direct and effect slot definition classes

15:15:39 kevingal usocket question: is there a way to configure a TCP socket so that calling READ-SEQUENCE on the socket stream reads in as many bytes as are available and then returns?

15:16:13 kevingal Right now, it's reading in a couple of bytes and then hanging because the buffer isn't full and there are no more bytes.

15:21:09 gilberth kevingal: There isn't.

15:23:11 kevingal Hm, do I need to read the bytes one at a time and keep checking usocket:socket-state, then?

15:23:17 gilberth And that would make READ-SEQUENCE not behave as specified by ANSI-CL. Your only chance is to to loop with using READ-CHAR-NO-HANG. READ-BYTE and LISTEN won't work in case of EOF (the connection being closed by the peer) because LISTEN would return NIL in that case.

15:24:40 xantoz_ ** NICK xantoz

15:25:06 gilberth I have no clue what SOCKET-STATE is. The proper API is LISTEN and as this is broken by specification READ-CHAR-NO-HANG.

15:26:52 gilberth I was working on a proposal to add an option to READ-SEQUENCE for UNIX behavior and/or better buffer access, but never managed to finish that.

15:27:41 kevingal Agh, annoying. I would need a READ-BYTE-NO-HANG.

15:27:52 gilberth As ANSI-CL is specified it's not possible to write fast TCP/IP servers and you can't do at all when you want to read octets.

15:28:09 gilberth kevingal: Right. That one is missing as well.

15:30:43 dnhester26 can anyone help? https://stackoverflow.com/questions/77942085/slot-definition-disappearing-when-using-two-metaclasses

15:31:04 dnhester26 _death: thanks for the help. Do you have any idea why the above is happening?

15:31:55 younder Just read sbcl/src/interpreter. As for fasteval it needs to be compiled in and SB-EXT:*EVALUATOR-MODE* needs to be :interpret. By default the variable isn't set, so evaluation isn't used.

15:32:17 younder I mean it is set to :compile

15:33:43 younder It was news to me as well. Just like that arena collector. SBCL just keeps throwing surprises at me :)

15:36:18 kevingal Thanks anyway, gilberth.

15:41:13 beach dnhester26: You probably need to define your methods specialized to oql-metaclass rather than to has-many-meta-class. Otherwise, it is not clear which method on direct/effective-slot-definition-class is going to be applicable. Yours or that of MITO.

15:42:02 beach dnhester26: Also, DEFMETHOD is a standard operator. There is no need to supply a CLOSER-MOP package prefix for it.

15:49:40 beach dnhester26: Oh, and MITO probably does the same thing that you do, which is to define its own direct/effective-slot-definition. So you probably need to define classes for those that have your slot-definition classes and those of MITO as superclasses, and have direct/effective-slot-definition-class return those class metaobjects.