Back to the changelog

• Improvement

  Work in the last phase counts

  An item no longer closes just because it reached the last phase. If that phase has tasks to complete, the item stays in progress until they are closed, exactly as in every other phase: before, it was considered finished on arrival, so the counts showed it as done and the recorded duration was shorter than the real one. Single-phase processes benefit too — the "log" case, where you record something that already happened but that still needs checking: these used to produce items that stayed active forever, with no way to close them.

• Improvement

  Placeholders stay reconciled

  When you reconciled "Tiziana" onto a real person, the replacement happened everywhere except on the assignee of a single task inside a phase — which is, in practice, the one that wins over all the others. The result was disorienting: the panel emptied out, and every new item on that process recreated the placeholder. That occurrence is now listed, counted and rewritten along with the rest, and the panel tells you where to find it by naming the process, the phase and the task. The "you are changing process design" warning appears in this case too, which is precisely the one that affects processes already published.

• Improvement

  A reconciliation list that tells the truth

  Empty values are gone from the list: they showed up as if they were assignees, with a Reconcile button that could not work. The same name typed with a leading space no longer counts as two separate placeholders. Processes involved are counted once rather than once per version of the same process. And when the list is partial — on large archives the scan stops at a threshold — it now says so, instead of looking complete.

• Security

  Internal errors no longer describe how the system is built

  When something failed, the message returned contained the technical error code, the name of the table involved and the failed request. The protection already existed, but because of an ordering problem at server start-up it applied to no call at all. With it, the other intended behaviour comes back too: when the database is overloaded the server says so and asks the client to retry, instead of returning a generic error that looks like a permanent fault.

• Security

  The reconciliation panel stays inside your organisation

  Two of the reads feeding that list were not filtering by organisation: names of people, processes and phases belonging to other customers on the same server appeared, and the counts came out inflated, typically double. Writing was already correctly confined, so no data was ever modified outside its own organisation: this was a read stepping outside the fence. An automatic check now prevents the problem from coming back.

• Improvement

  Sturdier contracts on titles, processes, templates and folders

  An item's title can no longer be wiped by accident: a value of the wrong type, easily sent by an automation, emptied it while the response said everything went fine. A process with several versions can now be deleted, whereas before the operation failed and the process stayed blocked forever with no explanation. Renaming a process onto a name already in use now says so and suggests an alternative, as creation already did. Templates and folders can no longer have duplicate names, and a folder can no longer end up inside itself by way of another one. Finally, saved templates carry the process KPIs with them: formulas used to be rewritten by hand. Templates saved before this version need regenerating to include them.