Abcl 20140920a without unnecessary commits. #44
Conversation
|
So isn't CFFI already broken for ABCL users? If we leave out the eval-when commit do things actually get worse for users with current ABCL? I can't remember if that's the case or not and don't really feel like reverting ABCL to test. |
|
On Tue, 2014-09-09 at 09:52 -0700, slyrus wrote:
As far as I can tell it's already (partially) broken, that's why I Stelian Ionescu a.k.a. fe[nl]ix |
|
Are we going to remove the HG cruft from the commits before this goes into the main cffi repo? |
|
On Tue, 2014-09-09 at 09:57 -0700, slyrus wrote:
I would like that very much. Stelian Ionescu a.k.a. fe[nl]ix |
|
Hmm... It seems to work on an ABCL even without the eval-when changes. I say we ditch HG cruft and the eval-when commit and get the rest of this committed. |
* call lispify-value on args to lisp callback functions, which requires passing the argument-types to covnert-args-to-lisp-values * clean up invocation of convert-args-to-lisp-values
* we get passed the actual library object, so no need to look it up in *loaded-libaries*
… value doesn't need to be lispified
…nside lispify-value
* but cheat and return nil in foreign-free because that's what it's supposed to return
use closures with variables to cache various methods to avoid a lot of introspection
* call java:jobject-lisp-value directly instead of lispify-value in copy-from-foreign-vector as it is much faster
e6cd20e to
dfe77de
Compare
|
As @slyrus said now without HG and the |
|
Awesome. Luis, can we get this committed soon? Thanks! |
|
Yes, I will take care of that tonight! |
As requested here new pull request with two unnecessary commits removed.