Daichi Tsuna authored 6 years ago and Kipp Cannon committed 6 years ago

Fixed bug in assertion error due to the fir_matrix changing its size. Made it to not change.
A bug in framecppchanneldemux (assertion error due to numsrcpads being 0) still persists, which appears when running the DAG but not when running in the head nodes.

Patrick Godwin authored 6 years ago and Kipp Cannon committed 6 years ago

gstlal_feature_listener: handle sqlite insert errors more gracefully, fix issue with reporting percent channels missed in packets

Aaron Viets authored 6 years ago and Kipp Cannon committed 6 years ago

gstlal_compute_strain:  Allow noise subtraction to be gated by channels other than ODC state vector.

- this code dates from a time when noise alone was what we "expected", now that statement is not true.  this patch changes the wording in some plots from "expected" to "noise model" to be more precise

chad.hanna authored 6 years ago and Kipp Cannon committed 6 years ago

inspiral.py: under max over coinc in gracedb upload since clustering has been added post coincidence

chad.hanna authored 6 years ago and Kipp Cannon committed 6 years ago

lloidhandler.py: 3 changes. 1) output the 'full' coinc trigger information to the kafka queue. 2) add a clustering keyword to the hander.  3) Fix a typo in gracedb verbosity message that caused a crash

gstlal-calibration:  Numerous changes to make output cleaned h(t) independent of start time, so that jobs run in parallel will produce smooth output without kinks.
lal_tdwhiten:  Added ability to synchronize data timestamps with filter updates.

 - gstlal_inspiral_pipe and gstlal_svd_bank take as option
--sample-rate
 - the waveforms that are generated with a lower fhigh
are shifted forward such that all waveforms in a given
sub bank have a common amount of early warning
 - this has the effect that the actual fhigh of waveforms
is less than requested in a given sub bank, since the
waveforms are shifted according to the waveform that takes
longest to go from fhigh to ISCO

Signed-off-by: Surabhi Sachdev <surabhi.sachdev@ligo.org>

- the SNR segments are stored with float boundaries, so be careful to
  convert things to floats when testing to see if they are covered by the
  segments to avoid repeated type casts inside the internal bisection
  search comparison operation

- reduce the operation count in appsink_new_buffer()

- when plugin cannot be loaded

- move more performance-critical methods into the C layer, this time the
  rich comparison methods

- recognizing that nothing uses the trigger class defined in
  snglinspiraltable except the trigger generator and its coincidence
  machinery, flatten the inheritence tree by one to improve attribute
  look-up performance