Benjamin Byholm
Might work. It is hard to tell without actually testing the finished code.
As always, in order of importance: lost messages, strangeness and performance. The proposal looks good at a glance. Lost messages and strangeness should be somewhat avoidable with regression tests. Performance...
Unfortunately Windows isn't the best OS for doing any development but .Net... Anyway, this sounds like the binding has problems loading libzmq. At the end of binding.cc is the Windows-only...
Good that you got it resolved. It at least used to be that local dlls always got picked up over system dlls, but that partly changed in the nearly fifteen...
Installing works perfectly fine. You just need to specify the `nodeddir` argument to npm, like the message says. Oh, right. It still won't build due to the v8 upgrades in...
Just figured I'd add a tip: http://www.appveyor.com/ offers free windows-based CI testing, like Travis does for real operating systems.
Turns out it is not that easy. First, a couple tests need updating. Second, the interceptors are not correctly handled in that they do require return values or will require...
Some tests are still failing related to the Holder change and interceptors. It might be that the tests simply need updating due to the functionality used not being allowed anymore....
I believe I fixed the tests. Let us wait and see.
@bnoordhuis Any chance you remember what the deal with `Holder` versus `This` in `node::ObjectWrap` was ten years ago and what it currently is wrt. to, say, the accessors tests here...