Loïc Mathieu

@yorickdevries subflow work exactly the same, if you have an output of type file it didn't return the file content but an internal storage file (which can be previewed in...

ForEachItem output returns a file with the merge of all the outputs of all subflows. If the outputs of the subflows are a FILE, it return a FILE with the...

@yuri1969 this is very interesting but now is not the ideal time to update 300 files ;) @Skraye has also a plan to integrate a linter soon, based on Spotless,...

In an execution with a lot of taskruns / attempts, this can add non-negligeable size to the execution context. To lower that, I propose to add only the workerId (we...

@tchiotludo `offset` is a PK so there is an index, but I agree it's stange. I need to dig a little as I think it is like that since the...

@tchiotludo this methods is used in tests to listen to the queue (mainly for execution or logs) and being able to receive messages to assert them. They do use the...

I overlooked, it's used in an other method in the queue that is used in a bunch of places: - FlowListeners - Worker - ExecutionController (to follow the execution) -...

@tchiotludo the best I can do easily is to add the consumerGroup and queueType condition to the fetch of the max offset, see https://github.com/kestra-io/kestra/pull/5834/files#diff-70145e96cdd165341d305afa24a65711a2819dba6297a2a93cacb0853c4228d8R224 If it's OK for you, I'll...

Another thing that comes to my mind is that from the ExecutionController and the LogController, when we follow executions/logs we will poll the DB with the configured poll frequency which...

It's only the main offset query that didn't contain the type, it then calls `receiveFetch` with that offset and it correctly have a condition on the type.