Your analysis looks sound and indeed we have to perform the task scheduling on a consistent set of input data: either we convert all input to UTC before scheduling, or we convert everything to a certain local time and then convert the output back to UTC before persisting into the database.
In any case it seems that have a working time that spans midnight is a possible situation, e.g. teams working in 3x8 shift pattern, so fixing Faces to support it properly is great!
Thanks for providing an in-depth analysis, a corresponding set of patches, and also for the upstream submission!
PS: the patches and related merge proposal still have to be reviewed, but currently the MP seems to contain unrelated changes, see my comment on it.
Hi Stuart,
Your analysis looks sound and indeed we have to perform the task scheduling on a consistent set of input data: either we convert all input to UTC before scheduling, or we convert everything to a certain local time and then convert the output back to UTC before persisting into the database.
In any case it seems that have a working time that spans midnight is a possible situation, e.g. teams working in 3x8 shift pattern, so fixing Faces to support it properly is great!
Thanks for providing an in-depth analysis, a corresponding set of patches, and also for the upstream submission!
PS: the patches and related merge proposal still have to be reviewed, but currently the MP seems to contain unrelated changes, see my comment on it.