Upgrading to Workflow involves many changes.  These few are basiciallynon-invasive and set the stage for what is to come.Introduced:workflow.html - a help page for the new idea of WorkflowCommitFest.STATUS_PARKEDNew Workflow Model to implement Workflow rules and functionalityEnforcement of the Following Constraints: - "At most one" non-Closed Commitfest Status Presence - "At most one" non-Moved POC Status Presence - Closed POC status include leavedate, otherwise it is absentAlso, added a convenience Make target for restoring test environmentAlso, added *.inc to the .editorconfig for html fragmentsWorkflow Highlights:Commitfests are now just Future-In Progress-ClosedYearly Close/Create at v18 Feature Freeze: Drafts v19, Bugs v18

In the interest of breaking up this major change to both UX, backend,and Workflow this patch separates out the main UX redesign choices forindependent consideration.There is a new workflow display that will be polished and describedonce the final components have been incorporated.  For now it simplyplaces all of the action menu items on the screen; and incorporatesa more coordinated and flow-oriented layout for the informationin the existing key-value table.The #history-table is hidden by default but does include potentiallyuseful information and so is expected to remain long-term.  The statushistory field in the keyvalue-table may be moved here at some point.The elements from the #keyvalue-table that did not get incorporatedinto the workflow display are placed into a new workflow supplementaltable.Unlike the key-value table, the new tables are horizontally compact.The #keyvalue-table is presently hidden but viewable.  Conceptually,the contents of this table should never be needed in this view andat some point the table itself can be removed.  For now it is leftas a fallback.The existing action bar has been renamed the administrative bar.This will be visible only to staff and has minimal logic for showingand hiding UI widgets.  That way, should the UI logic be faulty, itis generally still possible for staff to fix it.  In this commit thecontents of the bar remain unchanged, only the file name is different.

Remove all usage of close/next in the UI (just administrative actions).Remove all 'next' detection logic.  The UI will provide valid optionsto the user to choose between and they can make an informed choice;especially since for the typical user there will only be a single choice,to the Future Commitfest or back to the Parked one.  Committers canalso choose to bring any patch into the In Progress or Open/Interlude.The implicit constraint 'next' applied to the system was that,practically speaking, as Patches moved they always move to a Commitfestthat was started in the future: thus current_patch_on_commitfest couldjust use a descending start date sort to find the primary Commitfest.The transition action uses the Workflow implementation which allowsfor a Patch to re-enter the same Commitfest.  This changes theentrydate on the POC which is now used to find the current Commitfest.The Workflow module tries to improve upon the History messages beinggenerated.  Additional thoughts forthcoming in this area.  Should besimple enough to retain the status quo on these if desired, especiallysince I'm disliking the absence of a cfid column; which is why i addedthe relevant POC Commitfest name to the messages.The new code is a bit chatty in terms of feedback to the user.  Though,on the flip side, it intentionally doesn't feel the need to verifythat the button just clicked was intended.  That needs to be tidiedup based upon the guidelines I need to review.

Implement the Workflow rules checked for in the model within the GUI.This way, people only see the actions relevant to their situation.None of the URLs have changed and the only modifications to thecontext environment so far have been additive - thus up to herethere have been only a couple of minor backend changes.  Namely,enforcing the already implicit policy of only opening up at mostone of each Commitfest type at any given it.  And allowing forre-entrancy of a Patch to a Commitfest.The UI does enforce a "Commit or Reject" policy for Patch resolution.Staff can still use Returned with Feedback.  The author is expectedto Withdraw their patch if they do not plan to act on the feedback,or move it back to Parked and Waiting on Author.  Or they canWithdraw and create a new patch in the future.  While the Workflowis not against giving people options, the volume of both rejectedand returned resolutions is too small to warrant  distinguishing them.Hopefully the finished UX is intuitive.  The core three actors in theWorkflow - author, reviewer, committer - have their core actions incolumns left-to-right (though the columns are action-type oriented,not actor oriented).  The headers are the verbs for the actionsin the second row.  The third row is actor-oriented listing theauthor(s), reviewer(a), and committer along with related membershipactions.The fourth column is where the new movement implementation appears,making /close/next obsolete, especially with the addition of Parked.Having run out of actor categories the empty cell in the bottom-leftbecomes a perfect place to display the core information from thekey-value table (which has been hidden).  Namely, the status of theCommitfest and Patch this POC pertains to, the version (hyperlinkon the label to the Edit screen), the topic (also with a labelhyperlink to Edit), the Subscribe/Unsubscribe button, and thelast modified timestamp.Emails, CFBot, and Links also brought over in a key-value table;while History is also hidden but part of the final design.  Thehidden key-value table is being phased out.  The key elements fromit that need to be figured out are: - Status (is a form of History and that should get a re-work) - Created (doesn't seem actionable, especially now with Parked) - Latest Email (redundant with Email content?)ID doesn't seem to useful and can be gotten from the URL if needed.Title is already at the top of the page.All other fields are incorporated into the Workflow UI.

By creating the poc_enforce_maxoneoutcome_idx unique partial indexwe have ensured that every Patch has only one current commitfestdefined by having a status that isn't Moved.  Update Patch to findthis PoC via the index and then return its commitfest propertyfor current_commitfest (and note the method as being deprecated).

Make getting a POC while only having a patchid a single Workflow call.Patch and committer use the committer lookup and change interface.Status and close use the updatePOCStatus interface used by transition.All four leverage PoC as their main hook for patch and commitfest info.

It is desirable to be able to view the status history along withentry and leave dates.Add an edit button for authors to match up with reviewer and committer.Also, explicitly label what we are displaying in the final row.

The idea of having actions visible and then redirect for authenticationis acceptable in general but the /patch view is now driven by therole of the user viewing the page that it seems easier to simply showno actions if the user is not authenticated and add a note just abovethe action table indicating the read-only mode.Add missing is_staff protection for the administrative actions form.Just outright removing the key-value table if you are not logged in.Don't want to edit it, want to get rid of it.

Categories are labels/titles and should use title case.  Use initialcapital letter on all non-joining words.In Progress is an OK label choice but Ongoing communicates the samething with a single word just like all the other options.In the workflow the "Parked" status is being used for commitfestslabelled, e.g.,  Draft v18, so use Drafts as the category label too.Change Open to Bugs to relfect its new usage.Once committed a large technical refactor/rename pass over the modelsand code to make the enums match up with the usage is planned.

Authors and committers can move patches.Only open patches can be committed or rejcted.Closed patches can no be reverted/re-opened if committed/rejected.Only open patch can be moved.Enforce via triggers that future commitfests may not have patches.

JelteF left a comment

JelteF left a comment