To reproduce login to https://jupyter-jmatt.lsst.codes using your GitHub credentials.
Start the bash or terminal app using the drop down.
cat ./detected-sources.txt | sed -E -e 's/(True|False)/-/' > ./detected-sources-fixed.txt
cp ./detected-sources-fixed.txt ./detected-sources.txt
SourceDetectionTask should only add flags.negative if config.thresholdParity == "both"
v13.0 [Fall 2016] release
Try commit 3cc7631 in lsst_dm_stack_demo.
My best guess is it is due to conda-lsst's slightly older range limits on numpy (or similar). That is something I could test, but it requires rebuilding the conda binaries. So it'll take the rest of the night.
No. It's most definitely DM-903. If you use the commit I mentioned in the demo it will work fine. We are meant to put a tag on the demo to match the associated release. I assume DM-903 was merged after Frossie Economou created the release.
Excellent, checking out 3cc7631 works.
So it looks like it's DM-903 or similar.
I have made the 13.0 release of lsst_dm_stack_demo off this commit:
jmatt, please confirm it works for your binaries. I have tested it with the candidate source build on OSX and it seems fine.