libera/#commonlisp - IRC Chatlog

7:59:51 rumraisin ** NICK cayley5

9:37:20 _death it seems usocket has been broken on sbcl for almost a month now as it uses inaddr-any as a host to connect to, no matter the host provided? what am I missing here? https://github.com/usocket/usocket/commit/7a24c571541b1d046976f0c175f1f2658dbdf0f7#diff-eced9d709105db28d76a7b3d2182cb547fc815a4c1e37faa249df9d699b116a3R454

11:56:01 semz _death: afaict that branch is only taken if host is nil or eq to the wildcard host.

12:04:27 _death where do you see "host is nil"

12:14:46 jcowan _death: for local-domain (aka Unix-domain) sockets the host is always localhost, so the socket doesn't care what you specify

12:17:38 _death it's not about unix sockets.. it's just a bug that causes ordinary tcp socket-connect to fail, including everything that uses usocket for that, like drakma or dexador.. so I just reverted to 3006689e

12:18:25 semz _death: by reading the wrong line :S

12:19:00 semz it's local-host that'd have to be nil

13:40:14 JoshYoshi :trumpet:

13:40:16 JoshYoshi Hmm

13:40:19 JoshYoshi ** NICK Josh_2

13:40:42 Josh_2 :trumpet:

14:02:34 beach Is (DECLARE ((ARRAY FIXNUM) X)) valid syntax?

14:02:37 beach On the one hand, DECLARE takes a "declaration specifier" which is a list where the CAR is a "declaration identifier", and a "declaration identifier" is a symbol.

14:02:41 beach On the other hand, the dictionary entry for TYPE says that (typespec var*) is an abbreviation for (TYPE typespec var*) and as far as I can see, it doesn't restrict typespec to be a symbol.

14:02:42 beach Furthermore, SBCL accepts the example, which suggests that the SBCL maintainers considered the syntax valid.

14:03:47 beach A case for WSCL it seems.

14:04:04 pfdietz It is valid, IMO.

14:04:44 beach Because of the "on the other hand" argument?

14:05:10 thuna` It is convenient if valid but I would say that the more strict restriction applies so it's not valid

14:08:09 beach Oh, I guess I can ask people to run the example in the implementations at their disposal.

14:18:01 semz beach: http://www.lispworks.com/documentation/HyperSpec/Issues/iss349_w.htm seems to explicitly allow it.

14:19:50 pfdietz So, this was something that appeared in the final version, rather than in this draft of the standard?

14:20:01 beach semz: Thank you for pointing that out.

14:24:02 pfdietz Yes, ty.

14:26:11 beach Brucio-61: The s-expression-syntax library rejects this syntax. Does this fact reflect your choice of interpretation of the standard?

14:30:40 Brucio-61 beach: i simply haven't thought about that case before. i can probably adjust the syntax description to allow the compound type specifier variant

14:31:20 beach I see. Well, we should think about it at least. I wonder what other Common Lisp implementations do.

14:31:21 Brucio-61 i will probably have to perform some implicit conversion since the resulting node cannot use a compound type specifier as the declaration identifier (is that the term?)

14:31:44 beach Yes, declaration identifier.

14:47:44 jcowan I'd say that given a conflict between the main text and the glossary, the main text should normally win

14:50:48 beach There is also a conflict between two different sections of the main text.

14:51:48 beach But, given 3.3.3.1, I think the syntax is allowed.

14:51:54 beach ::clhs 3.3.3.1

14:51:54 Colleen Shorthand notation for type declarations http://www.lispworks.com/reference/HyperSpec/Body/03_cca.htm

14:56:03 jcowan Yes, that seems definitive

15:24:46 gilberth The trouble with that shorthand notation is that there is no way in ANSI-CL to tell whether a symbol names a type nor whether a symbol names a declaration. So you can't have macros that actually look at declarations. This may be troublesome when you want to implement a macro which binds variables serially, yet you want to use nested LETs. As you would need to pull out declarations applying to a specific symbol.

15:26:16 gilberth Consider e.g. that you want to implement DESTRUCTURING-BIND on your own. You would need to make it turn into just one LET. Can be done, but it may be awkward.

16:59:36 pfdietz gilberth: I don't understand that objection.  You can always recognize declaration forms, since they are conses who car is the symbol declare.

17:00:12 gilberth When I see (declare (foo x)) is that a type declaration or a FOO declaration?

17:01:13 gilberth Suppose I want to lift all type declarations for 'x', do I need to lift that declaration, or don't I need to do that?

18:10:13 _death but there's possibly a cltl2 way: (sb-cltl2:declaration-information 'declaration)

18:13:37 gilberth I was specifically speaking about ANSI-CL. I'm aware of that olde lexical environment access. And it's not SBCL-specific in any way. In practice I am fine with using implementation-specific means.

18:14:42 gilberth More severe IMHO is that neither EVAL nor COMPILE does take an environment argument. This makes it impossible to write a code walker.