GitLab

lalburst:  python 3 update

See merge request lscsoft/lalsuite!1662

lal:  python 3 updates

See merge request lscsoft/lalsuite!1661

XLALComputeAtomsForFmetric(): remove redundant, unused integral calculations

Closes #454

See merge request lscsoft/lalsuite!1649

- now that python 3 is required, some clean-up can be done
- remove "u" flags from string constants in XML I/O paths (all strings are unicode in python 3)
- remove "from __future__ import print_function" everywhere
- drop DictMixin compatibility code
- adopt the "yield from ..." construct
- update some neglected comments here and there

- now that python 3 is required, some clean-up can be done
- remove "u" flags from string constants in XML I/O paths (everything is unicode in python 3)
- remove __ne__() method from rate.Bins (automatically implemented using __eq__() in python 3)

lalapps_bucluster:  add a proof-of-life test

See merge request lscsoft/lalsuite!1653

remove duplicate `python_requires`

See merge request lscsoft/lalsuite!1657

A big cleanup of spec_avg.c code

See merge request lscsoft/lalsuite!1650

lal.utils.cache:

See merge request lscsoft/lalsuite!1654

- stop bitrot from making this program not work.

- change some "one liners" in an example so that they are, in fact, one-liners

don't use -q option for upgrade tests

See merge request lscsoft/lalsuite!1656

Removed a lot of commented out code, cleaned up the code for readability, and clarified comments
Fixed a serious bug that the SFTs were being normalized twice; this is probably not intended behaviour
Fixed a bug (or apparent bug) in the handling of gaps between SFT files for the spectrograms
Fixed a bug for the resolution of multiple bins averaged together for the sepctrograms
Code cleanup fixed some weird choices like a for loop not covering the last frequency bins of the SFT when making spectrograms (I suspect poor understanding from the last person editing the code)
Removed old Crab code that is not compiled anyway, and I suspect is no longer needed
Fixed a memory leak on program finish
Added test code for lalapps_spec_avg, though note that the test won't work on the previous version of the code because of several bugs that are fixed in this update. I'd like to consult with Greg Mendell before this code is approved/merged

this silences the output and the list of packages that are installed,
removed, upgraded, etc... can't be seen which makes checking that the
upgrade is working correctly difficult.

Remove python2 packages

See merge request lscsoft/lalsuite!1635

snglcoinc:  fixes

See merge request lscsoft/lalsuite!1652

remove use of glue's ProgressBar from lalsuite burst codes

See merge request lscsoft/lalsuite!1626

symptom:  string search fails to form triples from doubles approximately 50% of the time (50% of the triples are missing);  the inspiral search has no problem but the code is identical;  the only obvious difference between them is that the inspiral search is generally run with offset vectors containing only +ve offsets while the string search is run with +ve and -ve offsets.

I'm not exactly sure which of these issues was the culprit, but the following issues were found and fixed and the problem in the string search was fixed:

- when constructing doubles, construct and return all doubles containing events that can pass coincidence with an event < the boundary time, not just doubles with one of their own events < boundary time (I believe this was the only real bug in the code)
- that required adding a filter to the final output stage to remove coincs that do not contain at least one event < the requested boundary time.
- when assigning offset vectors to graph nodes, only normalize the entire offset vector do not normalize the components assigned to component nodes
- teach the singlesqueue's what their time slide offset is so they can report a t_coinc_complete that accounts for the offset (this wasn't a bug, it just makes the logic easier to understand)
- provide each singlesqueue with a coinc window unique to that instrument (this doesn't fix a bug, it addresses a FIXME to reduce latency)
- add lots of comments explaining how the algorithm works and change some variable names to help clarify what is happening at each stage

- pass +inf for t_complete instead of the time of the last event to guard against edge effects due to "<" vs "<=" issues in comparisons
- apply ntuple_copmare() inside the coinc engine using the existing feature for this.
- don't forget to pass --min-instruments to coinc engine

- allow the segments defining "on" times to be set on the command line
- make requesting a veto segment list that doesn't exist an error
- edit some documentation

Small update to tilts_at_infinity

See merge request lscsoft/lalsuite!1651