These are chat archives for ManageIQ/manageiq/rails5

16th
May 2016
Jason Frey
@Fryguy
May 16 2016 14:05
thought I'd ping this room with what I had in the ManageIQ room

@Fryguy
master is red, but I suspect that this commit from rails might be related: rails/rails@6b54316

note that there was a change to action_controller/metal/rescue.rb

in particular notice that index_of_handler_for_rescue was touched: https://github.com/rails/rails/commit/6b543166f7b0fd5712ffe831dad31d34f6de5dcd#diff-81f758fb88f738604ba750f9dde1f6bbL11

Chris Arcand
@chrisarcand
May 16 2016 14:05
I’m totally free, can dig in to this.
Jason Frey
@Fryguy
May 16 2016 14:05
@gtanzillo is testing now
but he can't recreate locally
Chris Arcand
@chrisarcand
May 16 2016 14:16
@gtanzillo @Fryguy I’ve got it locally.
Gregg Tanzillo
@gtanzillo
May 16 2016 14:20
@chrisarcand Are you seeing the failure locally? My tests are passing locally.
Chris Arcand
@chrisarcand
May 16 2016 14:20
I am, yes.
Oh my. The error is coming from an RSpec test request. This might be a breaking change to rspec-rails
Gregg Tanzillo
@gtanzillo
May 16 2016 14:22
Ah, that would make sense. Maybe I have an older spec.
Chris Arcand
@chrisarcand
May 16 2016 14:22
What specs are you running?
Gregg Tanzillo
@gtanzillo
May 16 2016 14:22
I have 3.5.0.beta3
First I run just the specific tests that failed then I ran all the request and controller testes:
rspec --seed 43054 ./spec/requests ./spec/controllers/
Chris Arcand
@chrisarcand
May 16 2016 14:24
:confused: Did you update Rails to the latest commit?
Gregg Tanzillo
@gtanzillo
May 16 2016 14:27
Yes, 6b543166f7b0. But don’t worry about it. If you recreated it then we’re good
Chris Arcand
@chrisarcand
May 16 2016 14:27
:+1:
Jason Frey
@Fryguy
May 16 2016 14:28
:+1:
If we narrow to down to Rails, you might want to bring in matthew or jeremy ( intentionally didn't @mention them yet ;) )
Chris Arcand
@chrisarcand
May 16 2016 14:29
Still double checking for one moment
Jason Frey
@Fryguy
May 16 2016 14:29
yeah
Chris Arcand
@chrisarcand
May 16 2016 15:00
Getting close. Short story is that wherever rails pulls in actionpack it’s using the (now old) rc release code. Not version by git commit like the other rails gems for some reason.
Jason Frey
@Fryguy
May 16 2016 15:02
o_O
Julian Cheal
@juliancheal
May 16 2016 15:03
Double o_O
Jason Frey
@Fryguy
May 16 2016 15:03
O_o o_O
Chris Arcand
@chrisarcand
May 16 2016 15:04
It’s our fault.
‘fault’.
I forgot we stick AR and AS to run bleeding edge. With this breaking change, we need to account for actionpack.
Fixing now.
Oleg Barenboim
@chessbyte
May 16 2016 15:11
why do we require those separately??
Chris Arcand
@chrisarcand
May 16 2016 15:12
For the same reason. rails/rails is just the meta gem. Setting that to the latest git commit, the related rails gems will still pull whatever the set semver allows (the cut release candidate)
Aka: release candidate isn’t bleeding edge enough for MIQ ;)
Oleg Barenboim
@chessbyte
May 16 2016 15:13
what??
Chris Arcand
@chrisarcand
May 16 2016 15:13
MIQ is currently using the latest commit on the rails meta gem, active record, and active support. But it uses the latest RC of actionpack.
Because a breaking change was introduced (after the rc) of actionpack, we need to use the latest of that as well.
I’m also going to switch us to the 5-0-0 branch of Rails, which will be closer to the actual 5.0.0 release than 5.0-stable.
(Matthew’s suggestion)
Oleg Barenboim
@chessbyte
May 16 2016 15:15
what is the difference between 5-0-0 and 5.0-stable?
Chris Arcand
@chrisarcand
May 16 2016 15:16
5-0-0 is the branch that 5.0.0 will actually be cut from. 5.0-stable continues through all patch releases (so, 5.0.1)
Joe Rafaniello
@jrafanie
May 16 2016 15:19
so, is the solution to add actionpack on 5.0-stable to our Gemfile?
Chris Arcand
@chrisarcand
May 16 2016 15:20
That introduces a few other version issues. I say run 5-0-0 instead (again, per matthews suggestion), which does not have this change introduced yet.
Nevermind, backwards. We should maybe switch to 5-0-0 but yes, let’s just pin actionpack @jrafanie
Works :ok_hand:
Oleg Barenboim
@chessbyte
May 16 2016 15:24
interesting methodology for branching -- @Fryguy would this be something that we should use?
Joe Rafaniello
@jrafanie
May 16 2016 15:25
:+1: dirty fix to get green is ok for now
Chris Arcand
@chrisarcand
May 16 2016 15:26
ManageIQ/manageiq#8732
Jason Frey
@Fryguy
May 16 2016 15:30
I'm am lost on the distinction between 5-0-stable and 5.0.0
is 5.0.0 off of 5-0-stable?
Chris Arcand
@chrisarcand
May 16 2016 15:30
They cut a separate branch for what will actually be contained in the 5.0.0 release, that’s all.
Yeah.
It’s deleted after the cut.
Jason Frey
@Fryguy
May 16 2016 15:31
are there commits in 5-0-0 that aren't in 5-0-stable?
Chris Arcand
@chrisarcand
May 16 2016 15:31
backwards, commits in stable that aren’t in 5-0-0 (yet)
Jason Frey
@Fryguy
May 16 2016 15:31
oh so for things that will make it to 5-0-1
Chris Arcand
@chrisarcand
May 16 2016 15:31
yeah.
Jason Frey
@Fryguy
May 16 2016 15:31
That is just confusing :)
So to answer @chessbyte 's question...on that alone I wouldn't adopt this branching strategy :)
Chris Arcand
@chrisarcand
May 16 2016 15:32
Mostly just for a clean cut, obv because it’s a patch version pretty much everything should make it into the cut from stable.
Yeah.
Oleg Barenboim
@chessbyte
May 16 2016 15:32
I think that is the same kind of control that our build team wants on what commits make it to GA vs errata
Jason Frey
@Fryguy
May 16 2016 15:32
yes it is
and it's exactly why I think it's confusing
because now there is a feature in "next"
and next time we cut an errata they probably won't want that either
I personally like the "leave the MR open forever, and if they want something merge it" style
anyway, @chrisarcand I see you have a PR which should fix the issue, but it leaves me with 2 questions
Chris Arcand
@chrisarcand
May 16 2016 15:35
Shoot
Jason Frey
@Fryguy
May 16 2016 15:35
1) Should we pin to 5-0-0 branch?
2) Is there an actual bug in the code that jeremy introduced that we need to make him aware of?
(Note that I still will merge what you have if it goes green...just thinking ahead)
Chris Arcand
@chrisarcand
May 16 2016 15:36
1) Prolly, but we can’t because of other version issues in our Gemfile. I’m looking at that right now (but we should merge my current PR for a green master anyway. Yup!)
2) I don’t think so. It’s just a case of pinning gems via git and meta gems.
Jason Frey
@Fryguy
May 16 2016 15:37
oh i see
Chris Arcand
@chrisarcand
May 16 2016 15:38
tbh I had just always assumed pinning the meta gem would pin the others. I had no idea we were running some latest RCs for gems and latest git commits on others. Tempted to just pin all of them as we’re technically using different points of rails master depending on the gem.
Oleg Barenboim
@chessbyte
May 16 2016 15:39
@chrisarcand yes, that is VERY confusing -- why are we even requiring specific gems from Rails if we are requiring Rails itself?
Jason Frey
@Fryguy
May 16 2016 15:39
so the ones in gems/pending/Gemfile are because we only use ActiveRecord and ActiveSupport over there
Chris Arcand
@chrisarcand
May 16 2016 15:40
When you require rails/rails on a specific commit, that specific commit just has a gemspec that specifies it needs a Rails 5 release of a gem. It doesn’t know to also use git. Meta gems :D Confusing but useful.
@Fryguy Ah true; I just put AP in there to be with the others. Want me to move?
Jason Frey
@Fryguy
May 16 2016 15:40
no it's fine for now...when you repin you can clean it up
Chris Arcand
@chrisarcand
May 16 2016 16:03
So, chatting more with Matthew, I think we should stick with 5.0-stable. 5.0.0 will be the ‘more stable’ of the two (despite the name), and presumably will be as close to the gem cut as possible, but that opens us to accidentally shipping a product with a branch that’s known to be deleted very soon (catastrophic). I’d rather just monitor 5.0-stable for us (which might not break anyway) than pin us to a branch that will break eventually. @Fryguy
Julian Cheal
@juliancheal
May 16 2016 16:05
So confusing. :+1: @chrisarcand
Joe Rafaniello
@jrafanie
May 16 2016 16:22
50 minute vmdb test suite is making me :sob:
Julian Cheal
@juliancheal
May 16 2016 16:32
Eep. It was 45 before :(
Chris Arcand
@chrisarcand
May 16 2016 16:57
@juliancheal You done with pluggable providers yet? :stuck_out_tongue_winking_eye:
Julian Cheal
@juliancheal
May 16 2016 16:58
@chrisarcand Just waiting for the tests to pass ;)
Chris Arcand
@chrisarcand
May 16 2016 16:58
We can break them into their own gem test suites, right?
Julian Cheal
@juliancheal
May 16 2016 16:58
@chrisarcand I wrote my message ^ on the phone. Didn’t realise it read like a :trollface: message. Didn’t mean it that way
That is the eventual plan yes :)
@chrisarcand I meant. Gah rails versions numbers is confusing. Good job Chris getting us all green again. :sparkles: :sparkles: :sparkles: :sparkles: :sparkles:
Chris Arcand
@chrisarcand
May 16 2016 17:02
haha
Joe Rafaniello
@jrafanie
May 16 2016 17:30

We can break them into their own gem test suites, right?

Maybe, @chrisarcand, I don't it's that easy though

I want to resurrect the parallel_tests thing if I can cleanly implement it ManageIQ/manageiq#5230
at least that will be with 1.5 cores on travis