It is sufficient to check the scripted diffs on one arch, i.e. CHECK_DOC==1.
Also, the default datadir should not be created by just running the tests.
It is sufficient to check the scripted diffs on one arch, i.e. CHECK_DOC==1.
Also, the default datadir should not be created by just running the tests.
Concept ACK regarding the assertion
Is it due to performance reasons we want to run scripted diff only for the arm-linux-gnueabihf build (CHECK_DOC=1)? One benefit of doing the scripted diff for all builds is to make it apparent from the build status overview that all builds fails. If only arm-linux-gnueabihf is red when a scripted diff fails it might look like a platform specific issue and hence require further investigation to see that it is a general scripted diff failure.
This argument could be raised for all checks that for performance reasons are specific to a given Travis build, but scripted diff should be fairly quick in the general case?
No need to run an arch independent check on each arch. If you feel like factoring out the CHECK_DOC into a different build stage (c.f. #11004 (comment)); Go for it!
@MarcoFalke Makes sense! Factoring out CHECK_DOC into a different build stage sounds like a good plan.
Concept ACK :-)
So without 5368614aa106b96e02ffb9e68eb86ebe197bc07f, it actually fails: https://travis-ci.org/MarcoFalke/bitcoin/jobs/272689632
Concept ACK, the Travis changes look right but I don't know enough about it to say for sure :)
Looks good to me. utACK fa40b0eb514e77928a13a007df170e5ede18ee3e
Thanks, will merge after the general issue with travis is resolved.
This is orthogonal to the travis issue and should not affect it.