0:01 AM
Blendify_
ericholscher, I see that that when running npm intall I get a bunch of deprecated warnings
0:03 AM
are these defined by a file in the repo? I am not finding the file if that is true
0:13 AM
ericholscher
0:15 AM
Blendify_
hmm do not think so. one of them is minimatch
0:15 AM
which is not found in a grep of the repo
0:16 AM
the others are win-spawn and graceful-fs
0:25 AM
ericholscher
might be dependencies
0:25 AM
npm loves dependencies
0:25 AM
agj would be the one to ask though, I don't fully know it :)
0:26 AM
could probably use a round of package upgrades
0:27 AM
humitos has quit
0:35 AM
Blendify_
ericholscher, I think it would make it a lot simpler if running grunt build should be part of the release process
0:35 AM
having to deal with merge conflicts with generated code is not that fun
0:38 AM
0:40 AM
agj
Blendify_: those are dependencies from our requriements
0:42 AM
is there any tooling around automating package releases in python? something like distzilla for perl.
0:42 AM
i think it makes sense as a release step, though we need to be on top of releases in that case
0:42 AM
ie - get in the habit of dev releases
0:43 AM
if we didn't have to put our password in the .travis.yml file, i'd say we could handle pypi and github releases there
0:43 AM
not sure how i feel about that
0:44 AM
there is also .gitattributes, but i only know of the feature, not sure if it would be applicable here
0:45 AM
ericholscher
travis has encryption stuff for this
0:46 AM
so you can encrypt your creds w/ their keys
0:46 AM
but yea, +1 on adding it as a release step and removing it from the repo :)
0:48 AM
Blendify_
is there a delay on the bot?
0:49 AM
agj
bot doesn't post for the theme here
0:50 AM
Blendify_
ah... That makes sense
0:50 AM
agj
ericholscher: yeah, i know of the encryption on keys, still bothers me posting creds to github :)
0:51 AM
probably not a huge issue if we really wanted that way though
0:51 AM
go proper ci
0:52 AM
Blendify_
ericholscher, is there a public number for the number of projects hosted on readthedocs?
0:52 AM
ericholscher
0:53 AM
Blendify_
cool thanks
1:09 AM
xafer has quit
1:12 AM
1:14 AM
cognifloyd has quit
1:15 AM
humitos joined the channel
2:18 AM
2:18 AM
as far as I know we are using up to date font awesome
2:30 AM
looks like we are using 4.6 and latest is 4.7
2:43 AM
ericholscher, should I uses branches on the main repo or make a fork?
2:54 AM
agj
Blendify_: either works
2:55 AM
Blendify_
agj, I think I will just keep it in the main repo
2:55 AM
agj
roger
2:55 AM
Blendify_
less work...
2:56 AM
agj
on #266, not sure it buys us anything but more work. +1 on closing that
2:57 AM
humitos: hey, missed your question the other day, my irc bouncer wasn't doing its job
2:57 AM
humitos: you still have questions?
2:57 AM
humitos
agj: hi, I'm about to go bed now
2:58 AM
let's talk tomorrow
2:59 AM
Blendify_ has quit
3:03 AM
agj
humitos: sure thing! talk to you tomorrow
4:19 AM
ericholscher
!m
4:36 AM
Blendify
Added some demo docs of things that have given me issues before
5:10 AM
ericholscher, around?
5:37 AM
ericholscher
Blendify: aye
5:39 AM
Blend|away
5:39 AM
Blendify
also what about adding an html logo to the demo docs?
5:40 AM
5:41 AM
although that might be too wide
5:41 AM
5:43 AM
ericholscher
hrm
5:44 AM
html logo seems fine on demo docs
5:44 AM
I'd like to better support that use case in general
5:44 AM
re: submoduleing -- that's tricky, we should probably not officially support it
5:44 AM
esp. if we're gonna switch to rendering the css on build, that kinda kills the submodule approach implicitly
6:02 AM
Blendify
ericholscher, do you have an opinion on which image I should include in the repo to use?
6:02 AM
ericholscher
the one from the blog looks good
6:07 AM
Blendify
ericholscher, the other one looks better actually...
6:08 AM
ericholscher
sure -- not a strong preference
6:09 AM
i'm about to head to bed -- but thanks again for your work!
6:11 AM
Blendify
o/ I should do the same considering you on the other side of the states then I
6:11 AM
Latter today I will do more bug fixes
6:21 AM
ericholscher
sounds good
6:23 AM
Blendify is now known as Blendify|zzz
6:53 AM
spm_draget joined the channel
8:46 AM
fasaxc joined the channel
9:32 AM
d0ugal has quit
9:36 AM
d0ugal joined the channel
9:41 AM
xafer joined the channel
9:48 AM
d0ugal has quit
9:50 AM
d0ugal joined the channel
12:54 PM
fasaxc joined the channel
17:37 PM
humitos: re conda: I _think_ all we need to do is rebuild the current image, which will update the conda version
17:38 PM
humitos
yeah, that would work
17:38 PM
because the docker image is downloading always the latest one
17:39 PM
ericholscher
yea
17:40 PM
humitos
but, are you thinking on re-build the 2.0 version? because the latest one has other changes also...
17:40 PM
so, the 2.0 being re-built should be called 2.1 :) haha
17:41 PM
ericholscher
heh
17:41 PM
I think agj is probably the best to answer that stuff -- I haven't had much hand in setting up the docker hub stuff to this point
17:44 PM
I triggered a rebuild of 2.0, which if that isn't using a cache, should hopefully fix it
17:47 PM
I believe 2.0 should be a tag, so it won't get the changes -- but should install the latest version of conda, because it isn't pinned
17:50 PM
17:50 PM
but it isn't showing any logs
17:54 PM
cognifloyd joined the channel
17:56 PM
agj
ericholscher: checked, you should have access to the docker hub org
17:56 PM
humitos
mmm... I'm not sure that docker pull readthedocs/build:2.0 will bring the new one, because it's the same tag
17:57 PM
agj
we can find a way to force it
17:57 PM
ericholscher
17:57 PM
a random thread I found said they run it with `--no-cache`, so it should rebuild the conda version to latest
17:57 PM
ericholscher runs to a meeting tho
17:57 PM
agj
yup
17:57 PM
we should be fixing this by pinning the installed version of conda instead, if anyone know of that
17:58 PM
this is all a symptom of installing from shell script
17:59 PM
18:00 PM
humitos
great!
18:00 PM
that should be the way of pinning the conda version in the docker image, then
18:00 PM
agj
aye
18:02 PM
humitos
taking advantage of this situation, I want to ask: what's the correct label for that specific issue? :)
18:03 PM
agj
Blend|away ericholscher: with building on release, we'd still have artifacts checked in, that won't change. too make folks pinning sphinx_rtd_theme to our repo instead
18:03 PM
humitos: which issue? i'm just getting caught up this morning before heading to the office
18:03 PM
humitos
18:03 PM
the one related to conda build and upgrading the image
18:04 PM
Blendify|zzz
my latest PR is ready for review
18:05 PM
Blendify|zzz is now known as Blendify
18:25 PM
agj
humitos: for something like that "operations" and "bug"
18:30 PM
18:30 PM
probably better as a png
18:35 PM
Blendify_ joined the channel
19:15 PM
humitos has quit
19:34 PM
ericholscher
bleh
19:34 PM
the rebuild still used installing: conda-4.3.11-py27_0 ...
19:35 PM
guessing we should patch the image to explicitly install conda
19:36 PM
agj: do we need to retag the builds?
19:36 PM
branch 2.0 into 2.1?
19:41 PM
Blendify_ has quit
19:41 PM
19:43 PM
Blendify
thanks agj
19:47 PM
humitos joined the channel
20:06 PM
Blendify_ joined the channel
20:19 PM
agj
ericholscher: you'll have to patch
rtfd/readthedocs.org with the build image name, in doc_environment/constants.py
20:19 PM
wonder if we should be using docker image metadata there actually, hrm
20:19 PM
for simplicity, we can reuse tags, that's fine
20:20 PM
ericholscher
I'm -1 on reusing tags, generally
20:20 PM
agj
not sure we need to get down to bug fix release minor versions on the images
20:20 PM
ericholscher
seems confusing
20:20 PM
agj
why?
20:20 PM
ericholscher
because its a tag, and if it changes over time, there isn't a way to know what version was what
20:21 PM
agj
for software packages i'm more strict, but we're one of the only consumers of this package
20:21 PM
20:22 PM
that is, each version of the image we release requires patching rtd currently
20:22 PM
the image metadata lookup needs to be a bit stronger
20:23 PM
docker image labels could be a way to solve this though, i hadn't thought of that
20:23 PM
inspect the image used for python support, instead of a dict lookup
20:23 PM
fastest solution is to reuse 2.0 tag though
20:23 PM
ooor
20:24 PM
change dict lookup to a version compare lookup, which fails for `latest` of course
20:24 PM
ericholscher
running to a lunch thing, but reusing 2.0 makes sense -- though I'm still not convinced we know what the actual issue is
20:24 PM
downgrading conda seems like the simplest fix, since it used to work