freenode/#sbcl - IRC Chatlog
Search
16:44:53
stassats
i doubt type derivation is ever going to be helpful, aside from maybe deriving the sign
16:45:32
|3b|
yeah, was wondering when it would fold without knowing the type directly, or wouldn't fold but would still have something to derive
18:14:10
flip214
stassats: I know, but back then it did work fine... I had that in my own code (that's how I found it!), and I tested afterwards
18:19:50
flip214
I'm 99.2% sure I _did_ test _that_ patch before submitting! I remember that I had 30 minutes uninterrupted hacking then ;)
18:20:25
flip214
and I changed that because I was confused about the duplicated slot that I couldn't find at first - because it was :included
18:21:13
flip214
I couldn't quickly find an example where the error _message_ gets tested, so I didn't write an explicit test for it
19:00:51
karlosz
i can't really find much in the spec. should (defun f (x) (the fixnum (g x))) only assert that the primary value of g is a fixnum?
19:03:27
Bike
well to check it you have to get all the values from g, then check the first one, and then return all the values. so it's sort of a pain
19:03:27
Colleen
Bike: drmeister said 41 minutes, 5 seconds ago: Can you check the log when you get in.
19:07:25
flip214
stassats: Please take a look at https://github.com/phmarek/sbcl/tree/pm - or do you want patches on the ML?
19:07:58
karlosz
OK well if its too complex to check we really shouldn't be inhibiting the tail call
19:08:17
karlosz
(the fixnum (g x)) is too complex to check, but the code generated is much worse than (g x)
19:08:45
Bike
"so its a bug to truncate the rest of the values" yeah CLHS THE is pretty clear on that
19:11:56
karlosz
thats not a good reason in my opinion because in lisp macros generate a lot of code that looks like that
19:41:58
karlosz
but its also true that our intuition about what is commonly encountered or not is sometimes wrong
19:44:36
flip214
stassats: I got a mail, "run failed...", " * 3b29cc42-b937-5096-7d15-7c4ff1852e1e failed (1 annotation)"
19:49:41
karlosz
stassats: https://github.com/sbcl/sbcl/blob/0769d503a349997afeedb6f6128f7e2af8789aa9/src/code/target-package.lisp#L743
19:50:12
karlosz
we use truly-the to force elision, but i think we should be able to tail call that without changing semantics, right?
20:28:42
stassats
(declare (explicit-check :result)) does what the truly-the is trying to do, with tail calls