freenode/#clasp - IRC Chatlog
Search
19:37:58
Bike
in this case it seems like the problem is %intrinsic-call is getting a function name that's not in the primitives list
19:39:07
Bike
ah. it's llvm.frameaddress.p0i8 instead of llvm.frameaddress now? is that an llvm 12 thing? whatever
19:58:51
drmeister
We had this problem with one of the intrinsics before and it seemed pretty arbitrary.
19:59:38
Colleen
frgo: drmeister said at 2021.01.28 17:08:40: Can you check that your boehmgc is installed using homebrew? That's what we use.
20:00:04
Bike
i also put something in so that if another intrinsic is changed i'll get a sensible error.
20:01:32
frgo
boehmgc is via homebrew. Compiled ok. Loading quicklisp via (load "quicklisp/setup.lisp") resulted in segmentation violation. Debugger backtrace says it's in a call to (OPEN #P"/Users/frgo/.cache/common-lisp/clasp-cclasp-boehm-0.4.2-3637-g057b30ba7-cst-macosx-x64/opt/quicklisp/quicklisp/package-tmpPS5HLO8R.fasp" :DIRECTION :IO :ELEMENT-TYPE :DEFAULT :EXTERNAL-FORMAT :UTF-8 :IF-EXISTS NIL :IF-DOES-NOT-EXIST :CREATE)
20:04:17
frgo
But that's not important. I am trying to figure this out myself. You are busy with proposal and stuff.
20:06:33
drmeister
Ok - thanks. What I can tell you is I installed BigSur, upgraded Xcode, cloned clasp 'master' and built it and then loaded quicklisp and it built fine.
20:06:59
drmeister
I can give the laptop to Martin and he could help match up library versions if it's something like that.
20:08:26
frgo
Oh - so once again a busy weekend... Yeah, me too. I actually miss writing proposal now that I no more in consulting but on the client side.
20:11:45
drmeister
Yes - the code-model message is meant to remind me to remove the code-model message.
20:12:38
Bike
it also only took fifteen minutes to build instead of twenty. dunno what's going on there
20:12:51
drmeister
We have up to date llvm-12 support. Now all we need is for llvm-12 to support us (linux/JITLink).
20:13:30
drmeister
The large code model slows down exception handling a lot. That might be part of it.
20:17:12
drmeister
The alternative was to go back to object files/dynamic libraries - but I wanted to really make sure that faso files worked in production.
21:26:15
Bike
drmeister: do you have any idea how we'd end up with characters with negative character codes? maybe we're misinterpreting words somewhere?
21:27:51
Bike
yeah. i mean, the problem is that DISASSEMBLE crashes slime because the string has these impossible characters in it
21:27:52
drmeister
There is a single function that generates character immediates as well as other immediates