Log file Viewer

[02:53:45] <tomprince> LiquidLight: It is broken and being investigated.

[02:54:48] <LiquidLight> good to know, then I'll stop spamming the refresh button for a bit :) thanks for the feedback

[03:25:02] <sigmavirus24> tomprince: do you know how to add TLS to meta.pycqa.org? It's just a CNAME to readthedocs

[04:13:34] <njs> dstufft: from backlog, are you writing a build system abstraction pep...?

[04:16:19] <tomprince> sigmavirus24: There is http://docs.readthedocs.org/en/latest/alternate_domains.html#cname-ssl

[04:26:46] <sigmavirus24> tomprince: I'm ashamed that I missed that

[04:29:13] <sigmavirus24> So I guess I could just run a server with let's encrypt and be done with that, eh?

[10:58:19] <dstufft> njs: yes

[18:26:38] <njs> dstufft: hmm. I was in the middle of cleaning up and updating my PEP to re-post as well...

[18:27:42] <dstufft> njs: competing PEPs aren't a bad thing :]

[18:28:19] <njs> dstufft: no, that was my theory. I just didn't realize we were going to take it to its logical conclusion :-)

[18:29:06] <dstufft> <Insert Oprah> you get a PEP, you get a PEP, everybody gets a PEP

[18:48:39] <njs> haha

[18:50:42] <njs> ...bit distracted today by this LIGO thing. The main figure uses matplotlib + a colormap that I helped design :-)

[18:57:11] <lifeless> dstufft: ping

[18:57:29] <dstufft> lifeless: pong

[18:57:58] <lifeless> dstufft: hows your new draft thing coming?

[18:58:20] <lifeless> dstufft: if its stalled, I'd like to arrange a brief hangout with you in a couple hours from now, to get a brain dump, so I can run with the idea

[18:58:37] <lifeless> dstufft: since the stalls on this thing overall are getting on my nether bits :)

[18:59:00] <dstufft> lifeless: good! been working on it today again other than when I had to join a call (and will stop for a bit since I have to run out to the store)

[18:59:11] <dstufft> I've switched over to workin on it though till we get some sort of a conclusion here

[18:59:15] <lifeless> dstufft: *separately*, can we remove --target from pip install ?

[19:00:51] <dstufft> lifeless: unsure. I know we use it in pip itself but it wouldn't be hard not to. beyond that I don't really know how widely it is used

[19:01:11] <lifeless> dstufft: just that we've got a PR open right now to deal with platlib not being copied

[19:01:23] <lifeless> dstufft: and looking at that xafer and I realised that it doesn't copy datafiles etc

[19:01:40] <dstufft> I wouldn't personally be majorly opposed to it, because it's sort of problematic since you're trying to take a bunch of dirs and flatten tem into one and it's not exactly possible to do it correctly

[19:03:12] <dstufft> the main reason I can come up with for keeping it is backwards compat

[19:04:37] <lifeless> ok

[19:04:42] <lifeless> so how about I mail distutils for input

[19:05:00] <lifeless> and if noone screams we deprecate it and wait for a few releases for use cases to be reported back ?

[19:06:39] <dstufft> lifeless: seems reasonable

[19:07:27] <lifeless> dstufft: what do we use --target for within pip ?

[19:08:05] <dstufft> lifeless: just for bundling shit in pip/_vendo

[19:09:59] <njs> lifeless: heya

[19:10:37] <lifeless> njs: hey

[19:14:51] <lifeless> ... and JSONLD.

[19:14:53] <lifeless> Of course.

[19:19:54] <njs> of course.

[19:20:36] <njs> I need to run for an hour or two, but would be around and interested if there's a hangout later

[19:29:20] <lifeless> njs: cool; it sounds like dstufft is about to send some interesting mail, once he gets back from the store :)

[19:48:37] <xafer> I think one question is whether we can do what --target does with --root/prefix ?

[22:21:34] <lifeless> xafer: the pr has an actual use case now

[23:05:08] <njs> ugh, the more I look at develop mode, the more of a mess it is :-(

[23:06:09] <njs> how is it supposed to handle purelib versus platlib?

[23:11:56] <njs> (also, does anyone know what the status is of PEP 491?)

[23:19:26] <xafer> good question, it is not present on https://www.pypa.io/en/latest/roadmap/

[23:34:37] <lifeless> njs: indeed, thats why I pushed back on pip owning develop mode right now

[23:34:42] <lifeless> njs: quagmire is quagmire

[23:34:50] <njs> I do see your point

[23:35:54] <njs> I mean, the way it *actually* works I think is that setuptools generates metadata, does an in-place build, and then picks some ~random directory and puts an .egg-link there

[23:36:16] <lifeless> fingers crossed

[23:36:17] <njs> and if we moved that last step into pip then we could at least make things less-broken later by fixing pip

[23:37:46] <njs> (last I checked flit refuses to try and understand .egg-link stuff, and does its own random thing involving symlinks, so it won't work on windows etc.)

[23:37:55] <lifeless> yay

[23:38:28] <njs> obviously fixing this for real will require evolving the interface at some point, and waiting for build backends to catch up and implement the new interface. but maybe it would be nice if we had some level to affect the old brokenness beyond "wait for build backends to catch up"?

[23:38:33] <njs> s/level/lever/

[23:39:59] <lifeless> well

[23:40:09] <lifeless> that would be nice

[23:40:21] <lifeless> perhaps 'works everywhere' would be such a lever

[23:40:56] <lifeless> dstufft: back from store?

[23:41:14] <njs> I have very mixed feelings. part of me is tempted to move just the .egg-link creation into pip, part of me is tempted to go with your solution of just leaving build backends to do whatever, and part of me is tempted to stay $#@ it, let's just leave the develop mode out entirely for now

[23:41:28] <njs> maybe that would at least encourage people to figure out how to make it work for real :-)

[23:41:53] <njs> (but I guess would be a serious impediment to uptake :-/)

[23:43:34] <lifeless> so right now, my thing leaves it out

[23:43:41] <lifeless> in the sense that its optional

[23:43:46] <lifeless> if you don't implement it, it will barf

[23:44:09] <njs> sure, that's always an option

[23:45:02] <lifeless> so flit can map develop -> their symlink thing

[23:45:26] <lifeless> setuptools_shim can map develop to setup.py develop

[23:53:00] <njs> I guess it also leaves room for folks to experiment with smarter develop modes. kinda.

[23:53:27] <njs> is it documented anywhere what interface pip actually uses with develop right now?I tried reading pip's code but it defeated me

[23:54:29] <lifeless> nope :)

[23:54:36] <njs> woooo