Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: daf_persistence, obs_base
-
Labels:None
-
Sprint:DB_W16_01
-
Team:SQuaRE
Description
While installing the stack on a new Ubuntu 14.04 box, using system Python, I encountered test/build problems that seem to indicate undeclare dependencies:
- daf_persistence requires a yaml python module
- daf_butlerUtils requires astropy.io
It was easy for me to work around these problems (with .deb packages and pip), so this is not a blocker for me, but I'm worried that we're accidentally subverting our dependency management system with some things provided by anaconda.
(Assigning to Tim Jenness for now since he seems likely to have been involved in any previous conversations or existing tickets on this subject, and hence may be able to quickly triage).
Attachments
Issue Links
- blocks
-
DM-4325 investigate distributing automated lsst_apps builds
- Done
- is blocked by
-
DM-4896 Add PyYAML EUPS package
- Done
- is triggered by
-
DM-3591 Add support for registry-free repository
- Done
-
DM-4171 Butler: change configuration from .paf to something else
- Done
- is triggering
-
RFC-125 Add pyyaml package to build
- Implemented
astropy was added in
DM-3591and it was noted in the pull request that this should be fixed, but only after the code had been merged. It's made worse because it's a dynamic import that only happens when the method is invoked. Nate Pease [X] is there a ticket for fixing this?Nate Pease [X] added the yaml dependency in
DM-4171in mid December. The new dependency on yaml was not spotted in the review. This is somewhat unfortunate as we seem to be relying on miniconda coming with yaml (lsstsw installs some items with conda during the deploy phase).Really we need an RFC on pyyaml adoption and to remove the astropy.io.fits code.