- Issue created by @spiderman
- 🇳🇱Netherlands johnv
Thanks you for your detailed report.
As youmay have noticed, 2.8 is not without error.
These days i Will try to fix the minimal use case, without access.
I'm filing this bug in case others are seeing similar behaviour, but I don't have a clear picture of what's involved yet. I'm working on a project with fairly complex use of Workflow module (including Workflow Access), as well as ECA and a variety of patches and newly-hatched contrib modules that are undoubtedly factors in the bug I'm seeing. I'll use this ticket to document my systematic testing to find a minimally reproducible version. Apologies if this is misplaced, but I'll take ownership to track this at least as far as identifying an actual bug (if any!)
These steps reproduce the problem in my project environment reliably, but I haven't yet validated they work in a vanilla Workflow site:
workflow_transition_history table).Again, I'm not clear which of the specifics of these steps are relevant as yet, and I'm confident that some constellation of extra pieces we have in play is creating this behaviour, but this is the most reliable set of steps I have currently.
I spent a fair bit of time yesterday in my debugger, tracing the form submission at the point when the Role2 user tries to edit the node for the first time (aside: dang! this got a *lot* harder to do in D8+). I was able to validate that the form data is submitted correctly, and seems to arrive at the NodeForm::save() in the $form_state variable as I'd expect.
However, it appears as though the form handlers "lose" the new version of the to_sid value somewhere along the way, and I gather just end up assuming no transition is required. That said, without knowing the Workflow code paths very intimately, I'm reluctant to keep spinning my wheels taking this approach. I figure it's time to try and strip down the problem to its barest essentials.
Our project includes several other modules that are certainly relevant, and many more that might tangentially be involved here. Having exhausted the obvious ways to troubleshoot in-situ, I plan to build up a failing scenario from scratch, slowly re-introducing some of the miscellaneous modules that are in play (roughly in order of how likely I think the are to be at issue):
Active
1.8
Code
Thanks you for your detailed report.
As youmay have noticed, 2.8 is not without error.
These days i Will try to fix the minimal use case, without access.