jni / gala

39%

master: 50%

LAST BUILD ON BRANCH py3k

branch: py3k

CHANGE BRANCH

x

Reset

• py3k
• batch-edges
• citation
• codacy
• conda-env
• cremi-io
• defaults
• docs
• easy-launch
• edt-seeds
• ev
• feature-cache-buffer
• fix-learning-mode
• fix-vtk
• float-watershed
• get-any-segmentation
• id-service
• learn-zmq
• master
• memreduce
• memreduce2
• minor-fixes
• nozeros
• purge-27
• pypi-annotations
• quick-features
• rtd
• rtd2
• speedup
• speedup-conflict
• tree-agglo
• tweaks

Build # 68

Build Type

push

travis-ci

Committed by jni

Commit Message

Test only partial merge history in Python 2

Due to apparent differences in floating point arithmetic between Linux
and OSX machines, testing the entire merge history in the second epoch
of training was working on my machines but not on Travis. I confirmed
this behaviour on the VLSCI Linux cluster, and checked how far I could
test the history for equality before failure. This is the measure used
in the test.

Run Details

1394 of 3568 relevant lines covered (39.07%)

0.39 hits per line