[Common] Protect against multiple downloads from TrackTuner

[mfaggin]

Possible solution to avoid any rewrite of the same files if the --pipeline option is used in Hyperloop for the device instantiating the TrackTuner

@ddobrigk @fcatalan92 @arossi81 @phymanshu

[github-actions]

O2 linter results: ❌ 21 errors, ⚠️ 0 warnings, 🔕 0 disabled

[ddobrigk]

Thanks @mfaggin ! I would remove this log with "FOUND" in favor of something after the loop reporting the count of devices, i.e. LOGF(info, "Number of track propagation tasks detected: %i", counter); or so. Could you please adjust?

[mfaggin]

Hi @ddobrigk , you are totally right. Done in my last commit

[ddobrigk]

One more thing: as you might have seen, the track propagation task is now deprecated and replaced by the propagationService, which has a series of autodetect improvements (including autodetection of needing or not needing the CovMat when propagating). It would be crucial, in fact, to have this solution also implemented in the file for the track propagation under Common/Tools, since the propagation service is an ideal task for pipelining. Could you please also make the necessary changes in the init function there? Thank you!

[mfaggin]

Hi @ddobrigk, actually the solution that I was proposing before was not working, since I was always getting N as number of devices, where N were the instances of track-propagation devices instantiated.

I changed the kogic directly in the TrackTuner utility, where before retrieving the file from CCDB it now checks if the desired file has been already downloaded. this should automatically fix the things also for the propagationService.

[mfaggin]

In this separate branch I put the implementation I had in mind to make the file being downloaded in different folders by the different TrackTuner folders: https://github.com/AliceO2Group/O2Physics/compare/master...mfaggin:O2Physics:fixCcdbQueryTrackTuner_gRandom?expand=1

The idea is to let the two instances downloade in the folders ./<number>. In this implementation <number> is just a random one extracted with gRandom. It would be better of course to use the device id, as I asked in Mattermost to @ktf , let's see if he has some suggestions on it.

[mfaggin]

Hi, I have just opened another draft PR here #12748 where I propose the usage of getfromTimeStamp instead of retrieveBlob, which should do what we want. This would imply the reload of all calibration files, but they are not too many (less than 10)

[github-actions]

This PR has not been updated in the last 30 days. Is it still needed? Unless further action is taken, it will be closed in 5 days.