David Rohr

Ideally, we should not have merge commits. Can you rebase without merge commits. Also, you should squash cleanups. If you don't intend to keep the history, please ignore, then we'll...

> Is the error in dataflow: > > ``` > Bulk-processing 2-prong Helix - Helix case > GPUassert: CUDA driver version is insufficient for CUDA runtime version /sw/SOURCES/O2/14918-slc9_x86-64/0/Common/DCAFitter/GPU/cuda/GPUInterface.cu 51 >...

If this is still causing trouble with the custom streamer, we could reconsider switching to a v2 object, and get rid of the streamer?

Just for my understanding, what are we trying to fix with this PR? Compatibility with new O2 and old ROOT? We went for this custom streamer since we said it...

Ok, but that means it is a fallout of the custom streamer. In general I absolutely fully agree to reduce dependencies, but I would do this independent from the streamer...

> @shahor02 this works in my synthetic tests (stage/bin/o2-testworkflows-early-forwarding -s --severity detail --early-forward-policy=always) . In the end I refactored the code to find the earliest spot where messages are guaranteed...

@ktf @singiamtel : I have no idea why this test suddenly fails in the dataflow CI: ``` The following tests FAILED: 29 - test_GPUsortCUDA (Failed) gpu Errors while running CTest...

Dataflow issue is independent, and compiles and runs on GPU, but should not be merged yet. Will comment more in the mail thread.

dataflow error is unrelated. From my side this could be merged, but should be checked with ITS experts before. I see also Ruben has already approved it.