libera/#clasp - IRC Chatlog

21:14:57 bike one of the frgo's PRs speeds up FASL loading by doing bigger file reads

21:15:25 bike it might help with cando too, but probably not as much

21:41:41 drmeister Oh nice.

21:42:07 drmeister Really nice. I didn't realize that was a bottleneck.

21:42:27 drmeister bike: Can you look at the new trampoline branch to consider merging it.

21:46:59 drmeister If you like the approach - hand coded trampoline & CIE and FDE sections for x86_64 and arm64 I can make it a PR

22:58:51 bike drmeister: oh i already filed a PR for it and wrote up a review, i just need to clean it up a bit before mreging

22:59:48 bike maybe do a release first as well

23:21:21 bike i'm thinking i'll try to get some of frgo's PRs in before releasing, since they're relatively simple and a FASL loading speedup and maybe the Mac fix would be nice

2:31:23 drmeister Yeah. Do frgo's first.

2:32:56 drmeister Can I catch a ERROR: #<FLOATING-POINT-INVALID-OPERATION @0x151df5a125a9> with a handler-case on arithmetic-error?

2:34:53 drmeister I know we have ext:with-float-traps-masked for masking floating point errors with NAN

2:42:27 drmeister What is `ext:float-nan-string` and `ext:float-infinity-string`

2:42:34 drmeister What are they supposed to do?

2:42:44 drmeister They are called but we don't have them defined.

2:50:31 drmeister If I recall correctly, we can't catch floating point errors because unix signals are shit

2:56:23 drmeister We are missing these functions...

3:33:41 bike you should be able to catch floating point invalid operation. ext:with-float-traps-masked prevents the error from being signaled at all

3:34:04 bike float-nan-string and float-infinity-string are supposed to be the printed versions of NaN and infinity

3:43:52 bike handling errors from a signal handler can be kind of weird, though, yeah