@shahor02 There is still one problem left which we are currently investigating which is connected to the timing of the calo bunches which we suspect is responsible for the crashes you see with the gamma2 raw fitter. Therefore the component is still not fully ready for a full test (but now it should be close). Please for the moment only merge the PR.

@mfasDa thanks, now I see factor 3-4 less cells then digits when creating from raw. But number of cells from the workflow

o2-emcal-reco-workflow --input-type digits --output-type cells --shm-segment-size 10000000000 --infile emcaldigits.root

is exactly the same as the number of digits. I guess this is not normal.
Also, there is still the problem of the links with no payload received, there should be no such links.

I am merging this PR w/o further CI checks since the changes are trivial.

@shahor02 3-4 digits per cell sounds reasonable as this matches the integration window of the zero-suppression (1 resample, 2 post samples over threshold).

There is still the issue with the time stamp we need to follow up - it could be that it currently at the wrong position of the bunch.

For digits to cells: Yes, this is reasonable only in the mode where we don't simulate the pulse (old run2 mode) - in the mode with pulse simulation (default) we need to treat them in the same way as raw data so they become either included in the RawToCellConverterSpec or we create a similar component as the RawToCellConverter for digits. Aside from the label handling where we have to consider how to assign labels to a bunch the porting should be straight forward.

@mfasDa I assume I was running digitns->cell in default mode when I saw 1:1 output. The cells from o2-emcal-reco-workflow should be exactly the same with digits or raw_from_digits input.
What about the links w/o any payload added?

@shahor02 Concerning the empty links: The equipment indexing was ported 1-1 from AliRoot for o2 and in AliRoot it had gaps:

• 2 empty equipments for EMCAL and DCAL 1/3 supermodules, which are connected to 1 SRU instead of 2
• 4 empty equipments for one additional DCAL 2/3 sector (2 DCAL supermodules connected to 2 SRUs), which were included in the initial stage of planning but then never realised.

Therefore the link indexing has gaps. As the equipment indexing is done differently in O2 we an get rid of the dead links however we need align this with the indexing in the hardware.

@mfasDa OK, as it is now, the emulated raw data will contain less links (since this which never get AddData call are not stored) than you register.

@shahor02 We will remove them, we just need to make sure first whether they are counted at hardware level in the FEEID in the RDH, because that definition must match between hardware, simulation and reconstruction as it is used in the inverse mapping for the tower ID calculation.