freenode/#clasp - IRC Chatlog
Search
19:14:42
drmeister
I would take https://github.com/clasp-developers/demo-clasp-cxx-interoperation and start by exposing your library and then make it an extension later.
19:15:24
drmeister
You can move it to an extension anytime - it's just some extra fiddly bits (wscript files) to deal with.
19:15:52
drmeister
I've taken pains to keep things general enough that others could create their own extensions.
19:16:42
drmeister
Extensions are more for C++ classes that you want to have managed completely by the GC. Extensions merge their class hierarchy into the clasp class hierarchy. Exposed C++ libraries don't need (or want) this.
19:18:48
drmeister
Bike: This thing with defclass at compile time - I'm about to reproduce the problem and then I'll get you a backtrace.
19:20:01
Bike
i'm not sure what we should do if that does turn out to be the issue. the standard is kind of weird about the point. we can work out something, though.
19:25:37
Bike
there was a discussion in #sicl here https://irclog.tymoon.eu/freenode/%23sicl?around=1550142743#1550142743 but not terribly actionable
19:36:05
drmeister
I just did a distclean and configure - will that update sicl? - It appears to have
21:27:10
Bike
if you do hit it again, probably turn off the parallel compiler, because that backtrace has nothing in it.
21:34:40
drmeister
frame 302 says it's in quicklisp/dists/quicklisp/software/babel-20171227-git/src/strings.lisp
21:36:39
drmeister
(defmacro with-checked-simple-vector (((v vector) (s start) (e end)) &body body) ...
21:42:45
drmeister
I changed both uses of 'e' (I have single letter variable names) to 'ebobo' - the compiler error message doesn't change - so I'm out of ideas.
21:44:55
drmeister
135: (CLEAVIR-COMPILE-FILE-FORM (DEFPARAMETER BABEL::*STRING-VECTOR-MAPPINGS* (BABEL-ENCODINGS:INSTANTIATE-CONCRETE-MAPPINGS :OCTET-SEQ-SETTER BABEL::UB-SET :OCTET-SEQ-GETTER BABEL::UB-GET :OCTET-SEQ-TYPE (SIMPLE-ARRAY (UNSIGNED-BYTE 8) (*)) :CODE-POINT-SEQ-SETTER BABEL::STRING-SET :CODE-POINT-SEQ-GETTER BABEL::STRING-GET :CODE-POINT-SEQ-TYPE BABEL:SIMPLE-UNICODE-STRING)))
21:45:52
Bike
instantiate-concrete-mappings is a macro that expands into a whole mess of stuff, as i remember.
22:05:53
drmeister
kpoeck: We just hit a problem with it - but this is after bike integrated some new compiler changes.
22:17:58
drmeister
Bike: Captain Hindsight says we should incorporate building cando into the final steps of compiler testing. It stresses the compiler in interesting ways.
22:57:35
drmeister
I just created a branch called 'work' that derives from testing and I'm cherry-picking the commits that follow your merge
0:28:49
drmeister
There is a bit of a problem with that - we have some things in the buildbot tied to dev. But we should probably change that rather than dev.
0:30:40
drmeister
Martin has a lot of buildbot scripts that are tied to dev he tells me. He thinks it's going to be a lot of trouble sorting this out.
1:02:37
drmeister
If we built the docker image and AWS machine images off of a particular branch - which should we use?
1:28:04
drmeister
Bike: To get back to the other question about defclass at compile time I got a backtrace with compile-file-serial
1:31:43
Bike
as for what we should do about it- we definitely want to make it known that it's a class at compile time, if nothing else for the find-class and typep optimizations
1:31:57
Bike
we don't have to do that by defining a class, though, we can just recorde it in the compilation environment somewhere
1:37:12
Bike
also, it's not actually because we share code, i introduced compile time defclass uh... a couple months ago i think