freenode/#clasp - IRC Chatlog
Search
12:48:22
Bike
beach: about storing docstrings in cleavir - in clasp we have enter instructions that store function names, and then the translate-to-llvm step specially load time value hoists it. i'm not sure how else to deal with it, since it's more a property of the enter than of the enclose even if enclose is what does the allocation- and if it's not a normal input to something it won't be hoisted normally.
12:51:49
attila_lendvai
drmeister: I stopped working on the build refactor for now. I'd fancy some quick feedback on it if you see anything objectionable: https://github.com/drmeister/clasp/pull/419
13:33:49
drmeister
frgo: Everyone here is excited about pybind11 - a lighter weight binding library between C++ and Python. https://github.com/pybind/pybind11
13:34:21
drmeister
It looks like boost.python where they stripped away a lot of boost dependencies and took advantage of C++11's variadic templates.
14:20:52
Bike
can't really reconstruct a lambda list suitable for debug from the enter-instruction's
14:30:48
Bike
drmeister: why does init.lsp make a bunch of compiler variables special? stuff like *clang-bin* that isn't used until bclasp
14:58:11
beach
Bike: Sorry, was busy making lunch for my favorite coauthor and then working in the afternoon.
14:59:04
beach
I haven't really thought much about docstrings. Is there anything you want me to give some thought? Or perhaps you have some direct question?
15:00:36
Bike
Well basically I want cleavir to be able to compile functions with docstrings. It can't right now, it discards them.
15:04:32
Bike
I think it makes sense to store it in the enter instruction. Then when an enclose is being translated it can grab that docstring and do whatever it does.
15:12:30
beach
True that. And this problem is implementation specific. We need an implementation-specific way of compiling ASTs in LOAD-TIME-VALUEs.
15:16:58
jackdaniel
ecraven: that's part of c++ awesomeness (and why interoperability with it is so hard)
15:18:23
ecraven
stassats: well, there obviously is a sort of unofficial ABI, clang and GCC can interact, as far as I know
15:19:30
jackdaniel
but calling full-blown C++ function without C wrappers compiled with C++ compiler – futile task
15:21:37
Bike
as far as i understand, C++ manglers have to include the entire template language, which is pretty involved
15:21:44
ecraven
so this means that theoretically a C++ library compiled with one compiler cannot be linked with a program compiled with a different compiler?
15:22:35
ecraven
yea, just thinking, if CL were as wide-spread as C++, I might ask this question the other way around
15:56:30
Bike
gonna try rewriting how allocate instance works. right now it does a bunch of initialization stuff it shouldn't
18:34:17
attila_lendvai2
drmeister: hi! it's nothing urgent, but when you are in an appropriate phase you could look at the build refactor PR and see if you see anything objectionable. especially regarding the building of extensions/ that I couldn't test. the commit messages should be enough, unless you catch something fishy