Log file Viewer

[02:52:57] <lifeless> dstufft: https://gist.github.com/rbtcollins/666c12aec869237f7cf7

[02:53:06] <lifeless> dstufft: care to eyeball that for me before I post to distutils-sig?

[02:54:07] <dstufft> lifeless: looking

[02:55:22] <lifeless> dstufft: thanks!

[02:55:33] <lifeless> dstufft: I just added one paragraph near the end, may want to refresh

[02:58:39] <dstufft> lifeless: did we ever verify if there was a pure python YAML implementation with a single source for 2.x and 3.x?

[03:00:22] <lifeless> dstufft: I promised I'd make some evil thing for you so you don't need to think about it

[03:00:49] <lifeless> dstufft: [which is roughly shove it down one dir deeper and do a __path__ patch at __import__ time

[03:00:59] <dstufft> oh, right

[03:01:00] <lifeless> dstufft: I haven't demonstrated that yet, but soon

[03:01:52] <lifeless> dstufft: I don't want to wedge on the syntax there; there are lots of ways we could go forward there, its the least interesting of the things to think about IMO

[03:02:14] <dstufft> oh yea syntax is absolutely the least interesting thing

[03:02:22] <dstufft> I just couldn't remember what happened with that

[03:02:50] <lifeless> yeah, its on me

[03:03:01] <lifeless> btw my evil testing hack ran into impl issues - its back on the back burner

[03:03:30] <lifeless> I think I can probably get something together, but its not the enabler stuff I need to prioritise atm

[03:14:53] <qwcode> seems like distutils-sig is suffering from mail delays

[03:15:06] <lifeless> dstufft: I'm going to prep a mail to distutils-sig

[03:15:26] <dstufft> lifeless: writing a quick note aobut it

[03:16:02] <dstufft> qwcode: https://status.python.org/incidents/l1jd6jcswlkh

[03:19:42] <qwcode> lifeless, surprised by your pep include an interface for "develop" and "install"... assumed it would just cover building wheels

[03:20:27] <lifeless> qwcode: pip install -e cannot work with just wheels

[03:20:47] <dstufft> install should die though.

[03:21:02] <lifeless> qwcode: if we don't include develop, then we don't enable other build systems to do -e

[03:21:36] <lifeless> dstufft: possibly

[03:21:44] <lifeless> dstufft: I'm sure we'll debate on list :)

[03:22:18] <qwcode> lifeless, yea, I see your pt.... I'm stuck in setuptools tunnel vision

[03:23:13] <lifeless> we probably should make both of them optional. I'll do that

[03:24:35] <lifeless> ok, decamping because private board stuff

[03:24:38] <lifeless> will sync up in a bit

[03:25:50] <qwcode> lifeless, can this pep really exist w/o also doing the sdist pep?, since it assumes an sdist will have this yaml file

[03:26:55] <dstufft> https://bpaste.net/show/bf6cc83d843f

[03:27:00] <dstufft> lifeless: ^^

[03:27:24] <dstufft> I'm still not sure I'm comfofrtable with the scope of things, the network effect is going to be unpleasent I think

[03:29:39] <dstufft> I gotta go, my daughter's dog is sick and needs to go to the emergency vet I think

[04:01:31] <qwcode> lifeless, "executing processes rather than loading Python modules, because tools like pip will often interact with many different versions". a process interface is fine I guess, but just because the api might be python, doesn't mean you'd try to load them all up in the main pip process (for the reasons you state), it does mean pip would need to grow a new script/callable for loading/running the api calls per build

[04:08:07] <qwcode> the idea that any/every build system could implement their own way to "install" and "develop"... that doesn't sound right... thinking "develop" installs need a PEP, and "install" should be removed

[05:42:22] <lifeless> qwcode: this is an alternative to the sdist spec, which was fundamentally confused about the layering

[05:42:31] <lifeless> qwcode: we sorted that out but haven't redrafted it

[05:42:45] <lifeless> dstufft: re: scope - ok, will add some analysis of that

[05:43:31] <lifeless> qwcode: develop probably needs a PEP yes, but I think as a first iteration having it be optional and distinct is better than nothing

[06:04:21] <lifeless> dstufft: provided-by is about efficiency, and avoiding pathological cases where build tools are nondeterministic

[08:39:13] <ronny> lifeless: still awake?

[17:21:37] <qwcode> the python mail delays are really frustrating : ( seeing discussions go by in the archive that I'm not seeing....

[17:23:14] <Wooble> I don't watch the archive, but I've definitely seen replies to messages I never saw, and then the original many hours later...

[17:23:41] <qwcode> for example. I got this, but neither of the previous emails in the thread today https://mail.python.org/pipermail/distutils-sig/2015-October/027363.html

[17:24:22] <qwcode> some emails I sent days ago, haven't made it to the archive

[17:24:47] <Wooble> Apprently I'm missing the last 13 python-dev messages. :/

[17:38:47] <dstufft> reminder: The entire mail.python.org was disabled for 3 days or so (hardware failure) and it's still attempting to process a pretty huge backlog