On 07/09/16 19:12, Ryan Day wrote:
On Wed, Sep 7, 2016 at 8:44 AM, Antti Kantee <[email protected]> wrote:
Ignoring hello_world(), I'd expect that program to output:
12
56
34
What is it actually printing?
That is indeed what it was printing. I wasn't paying enough attention to
the example and thought it might have crashed. But it seems to have worked.
Great, and even more so since the experiment was independently repeated
by Tim.
As for why people wouldn't want it, I don't really have a good answer other
than speculation. Greenlets is additional code, but not much. Greenlets is
pretty narrow in scope, so it isn't affecting other libraries. I think it
could be a default option without any obvious drawbacks.
Then my "vote" is to make it so, though it's of course your call as the
python maintainer.
When deciding whether to include Greenlets by default, I'm thinking more
abstractly about rumprun packages in general. Should packages inherit the
principles of the rump kernel? Users can add what they need, and leave out
what they don't. How can we make that easy to use from a user perspective?
Getting the balance right between usability and granularity is not easy.
It's one of those things I'd not worry too much about at least until
you have a use case which you can utilize to gauge the results of
componentization.
Additionally, you should consider that with rump kernels I had some
amount of direct control with the upstream sources, which did require
some modifications. If you run into a situation where you need to patch
upstream python, changes might not be worth it unless you can be certain
upstream accepts your patches.
Specifically for ./configure --with/out, I'd just pick a reasonable set
of defaults, and see where it leads.
I'm probably thinking too much about this one simple addon.... but these
chef scripts are taking forever to finish running :)
Every day I question if I'm underthinking or overthinking things ...
