dave__ has quit [Remote host closed the connection]
rdubya has joined #jruby
dave__ has joined #jruby
enebo has joined #jruby
dave__ has quit [Ping timeout: 268 seconds]
rdubya has quit [Quit: Leaving.]
<kares>
enebo: hey! I am close to a CI setup but I'll be using separate targets than you guys are (e.g. rails:test_sqlite3)
<kares>
just so that we run the Rails suite but use AR-JDBC's bundle dependencies (for CI/Sanity's sake)
<enebo>
kares did you see my change to rails:test on master the other day
<enebo>
kares: I think it properly invokes rails tests now since it uses the shell to change directories
<enebo>
kares: but rdubya, headius, and I were talking and we think for rails test we just do a shallow clone and put it in a subdir
<enebo>
kares: we are all at rubyconf
<kares>
enebo: good for you all
<kares>
no I do not like the approach since I would need to double git fetch and bundle install twice on CI
<enebo>
kares: but I think that invocation change and shallow clone will allow rails to test as a fairly quick clone
bbrowning is now known as bbrowning_away
<enebo>
kares: yeah my only issue is how to do test against HEAD of a branch of Rails?
<kares>
with a simple trick we can accomplish the same - by using gem 'activerecord', branch: '5-0-stable'
<kares>
that's what I am doing now
<enebo>
kares: ah yeah I sort of realized it after it came out of my head :)
<kares>
its either that or changing their Gemfile
<kares>
since they declare AR-JDBC it will always have its fallbacks
<enebo>
kares: I am also mildly conflicted on proper choice of that gem too
<enebo>
kares: for 50.x atm we are still not done getting 5.0.6 working AND their 5-0--stable is also broken atm
<kares>
simply - need to be in control and use our BUNDLE_GEMFILE when running their suite :)
<kares>
okey
<kares>
yeah I already noticed an issue with sqlite3]
<enebo>
kares: obviously at some point we will just track a branches HEAD but I was thinking about how to decide when to do a toggle
<enebo>
kares: yeah I am ok with having a consistent set of gems when we run our tests
<enebo>
kares: and you want to keep run times down and not load it all twice
<kares>
will leave your way of running (rake tasks) in for a while
<kares>
but thay should get obsolete
<enebo>
kares: my way did not take into consideration of not loading things twice but it tried to address your concern of loading a proper set of gems (rails and not ours in that case)
<enebo>
kares: but I am hoping it properly loads minitest now
<enebo>
kares: (I have not checked yet though)
<kares>
yep it does .. at least for me
<enebo>
kares: I figured it must because tests always run the same now and I no longer hack our pom.rb
<enebo>
kares: but I just never printed out a version string
<kares>
idid :)
<enebo>
kares: cool
<enebo>
kares: so another JRuby task which I dread is wy Dir.chdir + rake sh seems to remember original pwd
<kares>
you guys go have a serious AR-JDBC release party :) !
<enebo>
kares: we may do a little hacking today on it...does that count?
<kares>
yep I am also not using Dir.chdir anymore
<kares>
well whatever suits your style ... :)
<enebo>
kares: so mysql passes all rails AR tests (well without running their subdir tests which I will get set up soon/today) but I hit an AR code bug because we are failing over to some alternate code path which calls quote with 2 params
<enebo>
kares: since mysql has no actual quote method in our source I know this is just broken in AR itself
dave__ has joined #jruby
<enebo>
BUT native mysql passes an if above that and does not call that quote method at all
<enebo>
so I am like why??? :)
<kares>
there is a native quote piece on our end I think
<enebo>
I would like to solve that today. It will either be a) JRuby bug b) some string match or weird detection in AR c) something less obvious to me
<enebo>
kares: yeah but mysql is not using it or shouldn't anyways
<enebo>
kares: it goes down AR stack and finds the AR version and it only has one param
<enebo>
adatper.quote and not adapter.connection.quote
<kares>
aaah
<enebo>
kares: also quote(value, column=nil) and not quote_string
<kares>
well that's a different story I guess :)
<kares>
not sure
<enebo>
kares: it is funny since I think they basically totally killed this 2 param version in 5.0 and then when someone added the backoff they did not know that or forgot
<enebo>
kares: but I have been wrong about many things in life so I might be about all of this :)
<enebo>
kares: It is something I want to figure out today
<enebo>
kares: it only took 3 lines to make mysql work on 5.1 though
<enebo>
kares: although as I said we are not running the AR mysql subdir tests due to how their tasks are made so I know there are more problems now like migrations with compat of [4.2] using 5.1s id type
<enebo>
kares: I am going to cherry pick that system fix to 50-stable for the time being until we get a different solution
ahorek has joined #jruby
ahorek has quit [Client Quit]
rdubya has joined #jruby
<kares>
kk
<kares>
will open a PR when I am done with setting up Rails test on CI - just so you guys notice
<kares>
will be for 50-stable and planning on merging over to master afterwards
enebo has quit [Quit: Leaving.]
shellac has quit [Max SendQ exceeded]
dave__ has quit [Remote host closed the connection]