freenode/#sbcl - IRC Chatlog
Search
15:54:04
stassats
i might just implement it this weekend, maybe not efficiently, but this is getting embarrassing
16:19:16
scymtym
stassats: i have been working on that on and off for some time. would you give me until tomorrow to prepare a proposal?
16:25:44
nyef``
... A closer approximation could be generated, of course. Including error bounds to cover for the variation in length of years due to leap days. Accounting for leap seconds seems a little over-the-top for granularity, though.
17:08:55
scymtym
stassats: did you see this: when cross-compiling checkgen.lisp in: DEFUN TYPE-TEST-COST: STYLE-WARNING: The function TYPE-TEST-COST is called by REDUCE with (LIST) but it accepts (CTYPE)? that seems to be based on the sequence type rather than the element type
17:11:32
scymtym
(:key (function-designator ((nth-arg 1)))) -> (:key (function-designator ((nth-arg 1 :sequence t))))?
17:20:40
nyef``
Basically, make it easier to guess what the intent is. ":sequence t" says to me "something special is going on, you might want to look up what that means", while "elements-of" says to me "this is a sequence, and we're going to be using its elements here".
18:02:19
nyef``
stassats: Well, I can see why you don't want to change how these argument types are represented. SYS:SRC;COMPILER;MACROS.LISP, MAKE-CALLABLE-MAP is *brutal*.
19:43:26
nyef``
stassats: While I'm here, (nth-arg 0 :key key) or (key-of (nth-arg 0))? And (key-of (elements-of (nth-arg 1))), or (elements-of (nth-arg 1) :key key), or (elements-of (nth-arg 1 :key key))?
19:52:56
nyef``
Hrm. I might also try changing (rest-args :sequence t) to (elements-of (each-rest-arg))?
21:21:36
nyef``
The point is to have more readable surface syntax for higher-order functions in DEFKNOWN.
21:31:01
nyef``
I think that it's an improvement, but it's possibly a marginal one. I don't know that I'll push this upstream at this point.
22:41:11
stassats
it's been some time since anything in quicklisp got broken by better type conflict detection
22:57:17
stassats
"WARNING: Computing just-done stamp in plan NIL for action (ASDF/LISP-ACTION:LOAD-OP i" is getting so annoying
23:00:11
stassats
since the asdf people say "not our problem, all these .asd files that worked for years are broken"
23:05:50
stassats
"DEPRECATED-FUNCTION-STYLE-WARNING: Using deprecated function ASDF/BACKWARD-INTERFACE:RUN-SHELL-COMMAND -- please update your code to use a newer API."
0:11:50
nyef``
Is the first level of deprecation in SBCL a full STYLE-WARNING, or does it start below that point?
0:15:42
stassats
i think the next step for QUIT -> EXIT deprecation would be just to make QUIT call EXIT, not even going through the full WARNING stage
0:17:17
stassats
and i actually wouldn't have deprecated quit, it used to exit a thread? mark it as a bug and write a "minor incompatible change" in NEWS
0:20:25
stassats
quit should never go away in any case, and being stuck with a style-warning for eternity is not a good prospect
0:22:09
stassats
getting the API wrong twice, first, when originally designing it and then when deprecating it
0:45:33
stassats
part of the cunning plan "hey, sbcl deprecates things too" "what things?" sb-ext:quit" "not anymore!"
1:10:50
aeth
Get rid of quit and no one can leave SBCL anymore. A cunning plan for world domination.
1:40:51
stassats
thinking about specialized structure arrays, the hardest part would be deciding what to do when escaping a value
1:45:13
stassats
that greatly simplifies the implementation, just need to solve for interior pointers
1:52:38
nyef``
Include some flag and backpointer on the structure header to point to the array header.