#celery

ionelmc, nope… the problem still is there

I was starting with -P solo when I thought it was fixed

Verified, I've compiled the _billiard.so binary and I'm using github's celery/billiard at master

also, I'm using celery 3.1.0rc3

(also from master)

It's using _multiprocessing.Connection, which doesn't seem to have that attribute in python 2.7

…so I guess that "if is_pypy:" in billiard connection should also include python 2.x

Kronuz: make a bug report then

can't help you more :)

or you could try to fix it, can't be that hard to remove that if statement

Good morning

I'm using celery 3.1.0rc3, and after finding some problems with billiard.connection when I try to start a non-solo celery worker with multiple processes (it was saying _multiprocessing.Connection doesn't have setblocking attribute, under python 2.7)... I "fixed" that by also monkeypatching Connection to add setblocking (the same way it's done for pypy in the module billiard.connection). However...

I then encountered another (arguably bigger) error which segfaults 11 the whole thing when it tries to access the connection pool during the start up of the processes (this is in the billiard repository in GitHub, using master)

...I then went ahead and hunted the problem, it's definitely in the Connection object. So what I did was making billiard.connection use _connection3 for python3 and made a few minor adjustments in that module to make it python 2.7 compatible

That seemingly fixes all the problems for me now, I will make a pull request in the billiard repository for that...

Still, I have this question: is _billiard module not used in python 3? Are there any performance or otherwise disadvantages on using _connection3.Connection vs _billiard.Connection?

asksol: Are you around?