Details
-
Type:
Bug
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: log
-
Labels:
-
Story Points:1
-
Epic Link:
-
Sprint:DB_S19_02
-
Team:Data Access and Database
Description
The new changes that forward python log messages to lsst.log have resulted in pytest dumping a giant list of DEBUG messages related to matplotlib and/or font_manager after failed tests. This can require scrolling up through several pages of useless messages before you get to the actual pytest output.
Tim Jenness suggests either changing the default logging level we use when running pytest, or to configure our logger to ignore debug matplotlib messages.
Portion of an example "Captured log call" from a failed fgcm test:
------------------------------ Captured log call -------------------------------
|
font_manager.py 1343 DEBUG findfont: Matching :family=sans-serif:style=normal:variant=normal:weight=normal:stretch=normal:size=14.0 to DejaVu Sans ('/home/parejkoj/lsst/lsstsw/miniconda/envs/lsst-scipipe/lib/python3.6/site-packages/matplotlib/mpl-data/fonts/ttf/DejaVuSans.ttf') with score of 0.050000
|
font_manager.py 1343 DEBUG findfont: Matching :family=sans-serif:style=normal:variant=normal:weight=normal:stretch=normal:size=16.0 to DejaVu Sans ('/home/parejkoj/lsst/lsstsw/miniconda/envs/lsst-scipipe/lib/python3.6/site-packages/matplotlib/mpl-data/fonts/ttf/DejaVuSans.ttf') with score of 0.050000
|
font_manager.py 1343 DEBUG findfont: Matching :family=STIXGeneral:style=normal:variant=normal:weight=normal:stretch=normal:size=10.0 to STIXGeneral ('/home/parejkoj/lsst/lsstsw/miniconda/envs/lsst-scipipe/lib/python3.6/site-packages/matplotlib/mpl-data/fonts/ttf/STIXGeneral.ttf') with score of 0.050000
|
font_manager.py 1343 DEBUG findfont: Matching :family=STIXGeneral:style=italic:variant=normal:weight=normal:stretch=normal:size=10.0 to STIXGeneral ('/home/parejkoj/lsst/lsstsw/miniconda/envs/lsst-scipipe/lib/python3.6/site-packages/matplotlib/mpl-data/fonts/ttf/STIXGeneralItalic.ttf') with score of 0.050000
|
...
|
and more
|
...
|
Attachments
Activity
Great. I'm happy then. I had not realized that the -L options fixed up both root loggers.
Tim Jenness, thanks to K-T there is now more changes and one more PR (for log), could you look at it once again?
Looks great!
Updated docstrings for new class and merged all three PRs.
After messing around with all those pointless transformations I really wish we used Python logging everywhere on Python side and only used forwarding feature of lsst.log. I guess we'll be moving slowly in that direction during the next 10 years or so 
Well, if user configures logging through supported interface, that is providing a bunch of -L options to CmdLineTask (or pipetask), then both lsst.log loggers and their matching logging loggers will be set to the same level. If one says -L DEBUG then root loggers will be set to DEBUG so all debug messages will be seen (forwarded to lsst.log). If a child logger is configured (-L matplotlib=debug) then things will work too, debug messages sent to logging matplotlib logger will be accepted (because it is configured with DEBUG) and propagated to root logger (parent logger level does not matter when propagating) which will send them to handler(s). So I think with the mechanism implemented in this patch things will work as expected, potential trouble is when log4cxx is configured separately using some external mechanism (but this should not apply to CmdLineTask/pipetask because we always re-initialize log4cxx)