more kwargs? about pycares HOT 10 CLOSED

Hi there! Glad to hear you are using aiodns and pycares :-) Sure, such a patch would be welcome.

from pycares.

The patch is easy, but I am not having much luck installing and testing it. (I'm relatively new to python and python packaging.)

I modified the file pycares/_cfficore/init.py to have "class Channel" have the additional args.

I did "python ./setup.py install" and I can see it build c-ares and install some python code, but not including pycares/_cfficore

And when I "python tests/tests.py", it's clear that my code change to cfficore is not installed. Obviously the old one is installed somewhere, because the old tests all work.

from pycares.

untested code is in pull req #40

from pycares.

The PR will be built against different Python versions and operating systems, you can rely on the CI catching any mistakes :-)

from pycares.

I understand CI. My problem is modifying cfficore/_init_.py doesn't seem to cause anything to happen... including in the CI, which couldn't possibly be screwing up installation. If you look at pull req #40 you can see my change, and that my new tests fail as if I didn't change anything.

from pycares.

Right. So currently pycares is implemented kind-of twice. In C in src/ and using CFFI in cfficore. You need to make the change on both.

Currently all Python versions fail because it's not implemented in the C backend, and PyPy fails due to some type incompatibility. You can check the Travis output for all the details.

from pycares.

Thanks. That's worth a comment in the code :-) because I would never have figured that out from reading the Travis output! (although if I was more experienced with python I probably would have figured it out)

from pycares.

I am in an argument with ffi about the alignment of "struct in_addr *". As you can see from pull req #42 (which I created solely to get an additional test into CI) this problem was started by your change for issue 36 #36, which fails if there's a test that calls channel.set_local_ip().

I tried adding '...' and it didn't work, so I aligned struct in_addr like Linux does (uint32_t) and that works for both Linux and Windoze. Needless to say I'm a bit uncomfortable with this "solution", as an ffi newbie...

from pycares.

Yeah, I'll add a note on the README at least.

from pycares.

I have 2 pullreqs ready to merge, we've discussed them, they languish. Do you not get notifications for pullreq discussion?

from pycares.