2.x: Try fixing Travis CI lack of java

[akarnokd]

Travis CI has replaced the previous Precise image with the Trusty image as part of an ongoing rollout. Unfortunately, the Trusty image fails for oraclejdk7 because probably it is no longer supported (although I couldn't find anything about it).

This PR tries to identify what's available by trying a matrix of build setups.

[akarnokd]

Looks like Trusty only supports Java 8 builds:

There is also a test failure of MaybeFromRunnableTest.noErrorLoss due to being sensitive to delays.

Forcing dist: precise works.

[JakeWharton]

Why not move the build to JDK 8?

…

On Tue, Aug 1, 2017 at 4:27 PM David Karnok ***@***.***> wrote: Looks like Trusty only supports Java 8 builds: [image: image] <https://user-images.githubusercontent.com/1269832/28845459-639c07b8-7708-11e7-8a8d-10ffaefb4acf.png> There is also a test failure of MaybeFromRunnableTest.noErrorLoss due to being sensitive to delays. Forcing dist: precise works. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#5531 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEEEWTjqZfnSbLVMBFyonViJluDj9pGks5sT4obgaJpZM4OqOt3> .

[codecov]

Codecov Report

Merging #5531 into 2.x will increase coverage by 0.03%.
The diff coverage is n/a.

@@             Coverage Diff              @@
##                2.x    #5531      +/-   ##
============================================
+ Coverage     96.14%   96.18%   +0.03%     
- Complexity     5806     5822      +16     
============================================
  Files           631      631              
  Lines         41285    41405     +120     
  Branches       5732     5737       +5     
============================================
+ Hits          39693    39824     +131     
- Misses          629      630       +1     
+ Partials        963      951      -12

Impacted Files	Coverage Δ	Complexity Δ
...l/operators/observable/ObservableFlatMapMaybe.java	85.62% <0%> (-9.81%)	2% <0%> (ø)
...in/java/io/reactivex/subjects/BehaviorSubject.java	86.97% <0%> (-5.73%)	57% <0%> (ø)
...tivex/internal/schedulers/InstantPeriodicTask.java	61.11% <0%> (-5.56%)	7% <0%> (-2%)
.../operators/observable/ObservableFlatMapSingle.java	91.79% <0%> (-4.48%)	2% <0%> (ø)
...erators/completable/CompletableConcatIterable.java	95.91% <0%> (-4.09%)	2% <0%> (ø)
...nternal/operators/parallel/ParallelReduceFull.java	91.17% <0%> (-3.93%)	2% <0%> (ø)
...nternal/operators/observable/ObservableCreate.java	92.3% <0%> (-3.42%)	2% <0%> (ø)
...rnal/operators/flowable/FlowableFlatMapSingle.java	93.47% <0%> (-3.27%)	2% <0%> (ø)
...rnal/operators/observable/ObservableObserveOn.java	97.95% <0%> (-2.05%)	3% <0%> (ø)
...a/io/reactivex/internal/queue/MpscLinkedQueue.java	98.03% <0%> (-1.97%)	17% <0%> (-1%)
... and 76 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ac06c24...8752260. Read the comment docs.

[akarnokd]

It is an option; there was already a switch to oraclejdk7 back when the JDK 6 support dropped.

Should we switch to oraclejdk8 and run on trusty?

[JakeWharton]

That's what I use on all our projects which target Java 6 compatibility. We use JDK 8 and then Animal Sniffer to ensure no Java 8 APIs are referenced. It'll be nice when Java 9 comes out and its "-release 6" flag can be used to accomplish this.

…

On Tue, Aug 1, 2017 at 4:41 PM David Karnok ***@***.***> wrote: It is an option; there was already a switch to oraclejdk7 back when the JDK 6 support dropped. Should we switch to oraclejdk8 and run on trusty? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#5531 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEEERLUOx8rYuhszj2QPRSyOJD5xK_Iks5sT41ygaJpZM4OqOt3> .

[artem-zinnatullin]

I'd propose switching build to Docker where we can better control what environment is actually in the build and test against multiple versions of JDK properly (if needed).

Also with Docker we'll be able to reproduce CI builds very precisely (except hardware limits).

[akarnokd]

That's too risky and may not work due to the encrypted credentials we have here for maven upload.

[artem-zinnatullin]

I don't see problems with that, both files and environment variables can be easily passed to Docker container so that should solve maven upload.

And I actually think that keeping environment under Travis control (they update things) is more dangerous.