Rebecca Ewing authored 3 years ago and Patrick Godwin committed 3 years ago

gstlal-inspiral/python/dags/layers/inspiral.py: move delay_uploads option to upload section in config

line subtraction ringing test: fixed TF plotting.  lal_transferfunction:  working to add df and f0 properties

- fix timestamp for empty buffers when insufficient snr is available.

- make sure accesses to autocorrelation matrix are protected with the bank lock

- move the check for odd-lengthed autocorrelation vectors to the top to serve the dual-purpose of also checking for missing autocorrelation vectors

- in autocorrelation_matrix size functions, return 0 if the matrix is missing

in StreamBurca, build coinc row at once.  allows code to work with CoincTables classes that don't populate the coinc_burst table

- remove burca.StringCuspCoincTables.ntuple_comparefunc from coinc engine
- now in ranking statistic's .__call__()

This reverts commit be2c6e85.

inspiral_lr.py: now that 'real' triggers (triggers from forward templates) are identified 'signal_model', we need to specify this here.

lloidhandler.py: fix potential incorrect definition of 'buf_is_noise_model' and use '.search' property of the event to identify a 'signal model event' instead.

lloidhandler.py and streamthinca.py: fix bugs that cause 'noise_model' events leak into the trigger generator, and add assertion to make sure events are all from 'signal_model'.

runs every time.

Patrick Godwin authored 3 years ago and Patrick Godwin committed 3 years ago

merge Config min-instrument settings to only one is needed, ifos->instruments, min-ifos->min-instruments

- into src_pad_do_pending_events().  include logic to disable it so the heart beat code path still works.
- this is a step towards fixing the lack of stream ID handling
- take the opportunity to ensure every gst_event_push() operation in src_pad_do_pending_events() is properly error-checked.  there's still no error handling in the calling codes but that's a problem for another day.

far.py: lower number of zero-lag triggers needed for computing FARs by factor of 10 since trigger counter avoids counting triggers multiple times

increase timeout to grab triggers in trigger counter, make group ID unique for online jobs reading from Kafka