Change of behavior of om_gain: it used to apply weights

[ftadel]

The function om_gain used to return only one row per sensor, no matter how many integration points are used for these points. Now it returns one row per integration point.

In the example below, I have processed the same CTF example with 9 magnetometers (4 integration points) and 291 gradiometers (8 integration points) from Brainstorm with two versions of OpenMEEG:
https://www.dropbox.com/s/q1e9dfvalmz5hwo/openmeeg-2.1.zip?dl=0
https://www.dropbox.com/s/7m4y85ojxluk2a0/openmeeg-2.4.zip?dl=0

OpenMEEG 2.1: openmeeg_gain_meg.mat = [300 x 3009]
OpenMEEG 2.4: openmeeg_gain_meg.mat = [2364 x 3009]

This causes the current Brainstorm wrapper not to work anymore, as there is nothing to apply the weights after the call to om_gain:
https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/forward/bst_openmeeg.m#L458

And therefore users complaining:
https://neuroimage.usc.edu/forums/t/issue-forward-modelling-with-openmeeg-bem/10266/
https://neuroimage.usc.edu/forums/t/error-in-openmeeg-in-meg-data/8421

Is this related to this other issue?
#347

Did this become an optional parameter controlled with a command line flag?
Is it something that you will fix, or something that should be handled in the Brainstorm side instead? If so, what should we do: simply sum the rows multiplied by the weights?

[agramfort]

thanks @ftadel for the bug report... let us look into it ASAP.

[ftadel]

I confirm that replacing the sensor name from a simple index "1" to a text label "MEG001" fixes the issue:
brainstorm-tools/brainstorm3@7e463aa#diff-9e590703ce3a61a47c365d821b1164e6

[agramfort]

thanks