I want to be sure that I understand this issue. You would like to remove the code which calls getGaussianFlux, and simply do what is currently done in the "fixed" case: i.e., get the Centroid and Shape slots and call getFixedMomentsFlux.

And I assume that the slots are presumed to be set up prior to the construction of Inputs to the algorithm.

Simon, do you have time to review this issue? It is pretty simple and straightforward.

Please note that GaussianFlux now requires the shape slot to be setup before it can be run, which was not the case previously. However, the error handling when the Shape slot is not defined is not fatal. I will put a comment on this ticket to describe the work to be done.

When the shape slot is not setup, the creation of the Input object for the base_GaussianFlux plugins fails prior to the call to its measure() routine. To make error handleable, I had to switch two lines of code in sfm.py. This now produces the LogicError: key is not valid, but does not cause the task to completely fail, as it should. That will be corrected in DM-461, which I am now working on.

pgeepc2:/sandbox2/pgee/mylsst7/Linux64/mb12> git diff --stat master
include/lsst/meas/base/GaussianFlux.h | 17 ++++-------
include/lsst/meas/base/Inputs.h | 8 ++++--
python/lsst/meas/base/baseLib.i | 2 +-
python/lsst/meas/base/sfm.py | 2 +-
src/GaussianFlux.cc | 98 ++++++++++-----------------------------------------------------
tests/testGaussianFluxBasic.py | 5 ++--
6 files changed, 29 insertions, 103 deletions

Sorry, one more thing. Code is in meas_base, branch u/pgee/DM-1015

This is the diff based on the fixed and rebased version of the branch. It is ready for review.

include/lsst/meas/base/GaussianFlux.h | 17 ++++--------
include/lsst/meas/base/Inputs.h | 8 +++---
python/lsst/meas/base/baseLib.i | 2 +-
python/lsst/meas/base/sfm.py | 2 +-
src/GaussianFlux.cc | 135 ++++++++++++
tests/testGaussianFluxBasic.py | 12 ++++-----
6 files changed, 37 insertions, 139 deletions
pgeepc2:/sandbox2/pgee/mylsst7/Linux64/mb13>

Did this end up as a pull request on GitHub?

Review comments:

include/lsst/meas/base/GaussianFlux.h
Line 40 & 41: Is it possible to fix this documentation

O.K. Assuming you are running a branch build on buildbot before merging, this looks fine to me.

There are tickets in to fix these documentation issues.