Open gijzelaerr opened 5 years ago
what version of casacore and python-casacore?
input parameters and full log?
It used to run though the default test
note that this is using the unicode PR's for casacore and python-casascore.
input parameters are defaults, just running gocubical --data-ms test/SUBSET-D147-output.MS --model-list DATA --out-subtract-dirs 0
https://gist.github.com/gijzelaerr/8bd2bfe884d6157138da327938b7d8d9
ok can't track down the exact cause right now, but you need to set your solution intervals etc This fires the test on the CI server, see the accompanying parset https://github.com/ratt-ru/CubiCal/blob/master/test/d147_test.py https://github.com/ratt-ru/CubiCal/blob/master/test/d147-test.parset
docker run --rm -m 100g --cap-add sys_ptrace \
--memory-swap=-1 \
--shm-size=150g \
--rm=true \
--name=cubical$BUILD_NUMBER \
-v ${TEST_OUTPUT_DIR}:/workspace \
-v ${TEST_OUTPUT_DIR}:/root/tmp \
--entrypoint /bin/bash \
cubical:${BUILD_NUMBER} \
-c "cd /src/cubical && apt-get install -y git && pip install -r requirements.test.txt && nosetests --with-xunit --xunit-file /workspace/nosetests.xml test"
Think this is the same one I encountered: https://github.com/ratt-ru/CubiCal/issues/258
But why does this work on master? Same solution intervals etc. there...
This is with the latest casacore (3.1+), so I guess they became more strict in checking for flaky data.
These are with older gain tables (which can't be dumped outside of CASA, so there is a blank table inside the repo that is filled out with data). It may be that the new casacore cannot read these tables properly.
It is not a critical issue so if it breaks it breaks and we'll fix it later
While redoing the py3 porting work, I encountered this error when running
gocubical
ontest/SUBSET-D147-output.MS
both with py2 and py3:branch: https://github.com/ratt-ru/CubiCal/tree/py3_v2
note that this is using the unicode PR's for casacore and python-casascore.