libera/#commonlisp - IRC Chatlog
Search
3:17:56
aeth
easy bonus points to score: submit PRs to remove all (safety 0) except when the (safety 0) actually has a purpose (basically, to remove bounds checks) and even then you can probably move it to a tightly bound LOCALLY (instead of the typical function scope) to limit the chance of bugs surfacing there
9:14:16
_death
atgreen: but did you run your program with the unsafety declarations removed? then it might've exposed a bug
13:37:46
atgreen
SBCL team said "that's what you get" when you use (safety 0). The cl-postgres maintainer accepted my patch.
13:43:12
aeth
Is anything other than (speed 3) ever useful to write in code? I mean, code you give other people. Because (speed 3) sometimes makes the compiler go down another path (one that perhaps might be too slow to compile with when compiling on a 386 or whatever) that generates better code in a useful way...
13:44:27
aeth
And in SBCL it even gives you optimization notes (sometimes too many, if the optimization note is 'try dividing by another, nicer number') so it's kind of useful to keep contained to where it's needed instead of all over the program
13:55:53
aeth
But compilation-speed and space seem to be something that an implementation might reasonably not care about at all in 2022.
13:58:14
aeth
I wouldn't be surprised if some (safety 0) issues are from people who use other implementations and don't realize just how dangerous it is on implementations like SBCL that believe you when you say you want no safety.