Summary:
Depends on D20334. Ref T13272. After recent changes to make overheating queries throw by default, dashboard panels now fail into an error state when they overheat.
This is a big step up from the hard-coded homepage panels removed by D20333, but can be improved. Let these panels render partial results when they overheat and show a human-readable warning.
Test Plan: {F6314114}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13272
Differential Revision: https://secure.phabricator.com/D20335
Summary:
Depends on D20333. Ref T13272. Currently, dashboard query panels have an aesthetic but hard-to-see icon to view results, with no text label.
Instead, provide an easier-to-see button with a text label.
Test Plan: {F6314091}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13272
Differential Revision: https://secure.phabricator.com/D20334
Summary:
Ref T13272. Currently, the hard-coded default homepage looks like a dashboard but is actually rendered completely manually.
This means that various panel rendering improvements we'd like to make (including better "Show More" behavior and better handling of overheated queries) won't work on the home page naturally: we'd have to make these changes twice, once for dashboards and once for the home page.
Instead, build the home page out of real panels. This turns out to be significantly simpler (I think the backend part of panels/dashboards is mostly on solid footing, the frontend just needs some work).
Test Plan:
Loaded the default home page, saw a thing which looked the same as the old thing. Changes I know about / expect:
- The headers for these panels are no longer linked, but they weren't colorized before so the links were hard to find. I plan to improve panel behavior for "find/more" in a followup.
- I've removed the "follow us on twitter" default NUX if feed is empty, since this seems like unnecessary incidental complexity.
- (Internal exception behavior should be better, now.)
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13272
Differential Revision: https://secure.phabricator.com/D20333
Summary: Ref T13269. Make it visually more clear that the "Trigger" and "New Task / Edit / Bulk" dropdown menu items are buttons, not status icons or indicators of some kind.
Test Plan: {F6313872}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13269
Differential Revision: https://secure.phabricator.com/D20332
Summary:
Ref T13269. Currently, if you're on a milestone workboard like this:
> Projects > Parent > Milestone > Workboard
The "Parent" link goes to the parent profile. More often, I want it to go to the parent workboard. Try doing that? This is kind of one-off but I suspect it's a better rule.
Also, consolidate one billion manual constructions of "/board/" URIs.
Test Plan: Viewed a milestone workboard, clicked the parent link, ended up on the parent workboard.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13269
Differential Revision: https://secure.phabricator.com/D20331
Summary:
Ref T13269. These cards really have three states:
- Editable: shows a pencil icon edit button.
- You Do Not Have Permission To Edit This: shows a "no editing" icon in red.
- Hovecard: shouldn't show anything.
However, the "hovercard" and "no permission" states are currently the same state, so when I made the "no permission" state more obvious that made the hovercard go all weird.
Make these states explicitly separate states.
Test Plan:
Looked at a...
- Editable card on workboard: edit state.
- No permission card on workboard: no permission state.
- Any hovercard: "not editable, this is a hovercard" state.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13269
Differential Revision: https://secure.phabricator.com/D20330
Summary: If "GD" doesn't support a particular image type, applying a cover image currently goes through but no-ops. Fail it earlier in the process with a more specific error.
Test Plan: Without PNG support locally, dropped a PNG onto a card on a workboard. Got a more useful error.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20328
Summary: Ref T13269. I refactored this late in the game to organize things better and add table cells around stuff, and accidentally broke the relationship between the "Rule Type" selector and the value selector.
Test Plan: Switched rule type selector from "Change Status" to "Play Sound", saw secondary control update properly.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13269
Differential Revision: https://secure.phabricator.com/D20326
Summary: This is a copy/paste/find-and-replace-all of the status rule added by D20288.
Test Plan: Made some triggers, moved some tasks, edited some triggers. Grepped for the word "status" in the new file.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Differential Revision: https://secure.phabricator.com/D20325
Summary:
Depends on D20321. Fixes T12175. Ref T13074. Now that before/after PHIDs are suggestions, we can give the server a more complete view of what the client is trying to do so we're more likely to get a good outcome if the client view is out of date.
Instead of passing only the one directly adjacent card PHID, pass all the card PHIDs that the client thinks are in the same group.
(For gigantic columns with tens of thousands of tasks this might need some tweaking -- like, slice both lists down to 10 items -- but we can cross that bridge when we come to it.)
Test Plan:
- Dragged some cards around to top/bottom/middle positions, saw good positioning in all cases.
- In two windows, dragged stuff around on the same board. At least at first glance, conflicting simultaneous edits seemed to do reasonable things.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13074, T12175
Differential Revision: https://secure.phabricator.com/D20322
Summary:
Depends on D20320. Ref T12175. Ref T13074. Currently, when you move a card between columns, the internal transaction takes exactly one `afterPHID` or `beforePHID` and moves the card before or after the specified card.
This is a fairly strict interpretation and causes a number of practical issues, mostly because the user/client view of the board may be out of date and the card they're dragging before or after may no longer exist: another user might have moved or hidden it between the last client update and the current time.
In T13074, we also run into a more subtle issue where a card that incorrectly appears in multiple columns fatals when dropped before or after itself.
In all cases, a better behavior is just to complete the move and accept that the position may not end up exactly like the user specified. We could prompt the user instead:
> You tried to drop this card after card X, but that card has moved since you last loaded the board. Reload the board and try again.
...but this is pretty hostile and probably rarely/never what the user wants.
Instead, accept a list of before/after PHIDs and just try them until we find one that works, or accept a default position if none work. In essentially all cases, this means that the move "just works" like users expect it to instead of fataling in a confusing/disruptive/undesirable (but "technically correct") way.
(A followup will make the client JS send more beforePHIDs/afterPHIDs so this works more often.)
We could eventually add a "strict" mode in the API or something if there's some bot/API use case for precise behavior here, but I suspect none exist today or are (ever?) likely to exist in the future.
Test Plan:
- (T13074) Inserted two conflicting rows to put a card on two columns on the same board. Dropped one version of it underneath the other version. Before: confusing fatal. After: cards merge sensibly into one consistent card.
- (T12175) Opened two views of a board. Moved card A to a different column on the first view. On the second view, dropped card B under card A (still showing in the old column). Before: confusing fatal. After: card ended up in the right column in approximately the right place, very reasonably.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13074, T12175
Differential Revision: https://secure.phabricator.com/D20321
Summary:
Fixes T13270. In Diffusion, the "Code" tab is linked in a weird way that isn't consistent with the other tabs.
Particularly, if you navigate to `x/y/z/` and toggle between the "Branches" and "History" tabs (or other tabs), you keep your path. If you click "Code", you lose your path.
Instead, retain the path, so you can navigate somewhere and then toggle to/from the "Code" tab to get different views of the same path.
Test Plan: Browed into a repository, clicked "History", clicked "Code", ended up back in the place I started.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13270
Differential Revision: https://secure.phabricator.com/D20323
Summary:
Depends on D20308. Ref T5474. The element which previews what will happen when you drop a task somewhere can cover the bottom part of the rightmost column on a workboard.
To fix this, I'm trying to just fade it out if you put your cursor over it. I tried to do this in a simple way previously (":hover" + "opacity: 0.25") but it doesn't actually work because "pointer-events: none" stops ":hover" from working.
Instead, do this in Javascript. This is a little more complicated but: it works; and we can do the fade when you get //near// the element instead of actually over it, which feels a little better.
Test Plan:
- Shrank window to fairly small size so that the preview could cover up stuff on the workboard.
- Dragged a card toward the rightmost column.
- Before: drop action preview covered some workboard stuff.
- After: preview faded out as my cursor approached.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20320
Summary:
In some cases, we show a limited number of one type of object somewhere else, like "Recent Such-And-Such" or "Herald Rules Which Use This" or whatever.
We don't do a very good job of communicating that these are partial lists, or how to see all the results. Usually there's a button in the upper right, which is fine, but this could be better.
Add an explicit "more stuff" button that shows up where a pager would appear and makes it clear that (a) the list is partial; and (b) you can click the button to see everything.
Test Plan: {F6302793}
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Differential Revision: https://secure.phabricator.com/D20315
Summary:
Ref T5474. In 99% of cases, a separate "archived/active" status for triggers probably doesn't make much sense: there's not much reason to ever disable/archive a trigger explcitly, and the archival rule is really just "is this trigger used by anything?".
(The one reason I can think of to disable a trigger manually is because you want to put something in a column and skip trigger rules, but you can already do this from the task detail page anyway, and disabling the trigger globally is a bad way to accomplish this if it's in use by other columns.)
Instead of adding a separate "status", just track how many columns a trigger is used by and consider it "inactive" if it is not used by any active columns.
Test Plan: This is slightly hard to test exhaustively since you can't share a trigger across multiple columns right now, but: rebuild indexes, poked around the trigger list and trigger details, added/removed triggers.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20308
Summary:
Ref T5474. Allow columns to play a sound when tasks are dropped.
This is a little tricky because Safari has changed somewhat recently to require some gymnastics to play sounds when the user didn't explicitly click something. Preloading the sound on the first mouse interaction, then playing and immediately pausing it seems to work, though.
Test Plan: Added a trigger with 5 sounds. In Safari, Chrome, and Firefox, dropped a card into the column. In all browsers, heard a nice sequence of 5 sounds played one after the other.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20306
Summary:
Ref T13074. If you hit a prompt on a drop operation (today: MFA; in the future, maybe "add a comment" or "assign this task"), we currently leave the board in a bad semi-frozen state if you cancel the workflow by pressing "Cancel" on the dialog.
Instead, put things back the way they were.
Test Plan: Dragged an MFA-required card, cancelled the MFA prompt, got a functional board instead of a semi-frozen board I needed to reload.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13074
Differential Revision: https://secure.phabricator.com/D20305
Summary:
Ref T5474. The first rough cut of triggers showed some of the trigger rules in a tooltip when you hover over the "add/remove" trigger menu.
This isn't great since we don't have much room and it's a bit finnicky / hard to read.
Since we have a better way to show effects now in the drop preview, just use that instead. When you hover over the trigger menu, preview the trigger in the "drop effect" element, with a "Trigger: such-and-such" header.
Test Plan:
- This is pretty tough to screenshot.
- Hovered over menu, got a sensible preview of the trigger effects.
- Dragged a card over the menu, no preview.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20304
Summary: Ref T5474. When you view the main page for a rule, show what the rule does before you actually edit it.
Test Plan:
Viewed a real trigger, then faked invalid/unknown rules:
{F6300211}
{F6300212}
{F6300213}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20303
Summary:
Ref T5474. This provides a Herald-like UI for editing workboard trigger rules.
This probably has some missing pieces and doesn't actually save anything to the database yet, but the basics at least roughly work.
Test Plan: {F6299886}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20301
Summary:
Ref T10335. When you (for example) drag a "Resolved" task into a column with "Trigger: change status to resolved.", don't show a hint that the action will "Change status to resolved." since this isn't helpful and is somewhat confusing.
For now, the only visibility operator is "!=" since all current actions are simple field comparisons, but some actions in the future (like "add subscriber" or "remove project") might need other conditions.
Test Plan:
Dragged cards in ways that previously provided useless hints: move from column A to column B on a "Group by Priority" board; drag a resolved task to a "Trigger: change status to as resolved" column. Saw a more accurate preview in both cases.
Drags which actually cause effects still show the effects correctly.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T10335
Differential Revision: https://secure.phabricator.com/D20300
Summary:
Ref T10335. Ref T5474. When you drag-and-drop a card on a workboard, show a UI hint which lists all the things that the operation will do.
This shows: column moves; changes because of dragging a card to a different header; and changes which will be caused by triggers.
Not implemented here:
- Actions are currently shown even if they have no effect. For example, if you drag a "Normal" task to a different column, it says "Change priority to Normal.". I plan to hide actions which have no effect, but figuring this out is a little bit tricky.
- I'd like to make "trigger effects" vs "non-trigger effects" a little more clear in the future, probably.
Test Plan:
Dragged stuff between columns and headers, and into columns with triggers. Got appropriate preview text hints previewing what the action would do in the UI.
(This is tricky to take a screenshot of since it only shows up while the mouse cursor is down.)
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T10335, T5474
Differential Revision: https://secure.phabricator.com/D20299
Summary: Depends on D20287. Ref T5474. This hard-codes a storage value for every trigger, with a "Change status to <default closed status>" rule and two bogus rules. Rules may now apply transactions when cards are dropped.
Test Plan: Dragged cards to a column with a trigger, saw them close.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20288
Summary:
Depends on D20286. Ref T5474. Attaches triggers to columns and makes "Remove Trigger" work.
(There's no "pick an existing named trigger from a list" UI yet, but I plan to add that at some point.)
Test Plan: Attached and removed triggers, saw column UI update appropriately.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20287
Summary: Depends on D20279. Ref T5474. Modernize these transactions before I add a new "TriggerTransaction" for setting triggers.
Test Plan: Created a column. Edited a column name and point limit. Hid and un-hid a column. Grepped for removed symbols.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20286
Summary:
Depends on D20278. Ref T5474. This change creates some new empty objects that do nothing, and some new views for looking at those objects. There's no actual useful behavior yet.
The "Edit" controller is custom instead of being driven by "EditEngine" because I expect it to be a Herald-style "add new rules" UI, and EditEngine isn't a clean match for those today (although maybe I'll try to move it over).
The general idea here is:
- Triggers are "real" objects with a real PHID.
- Each trigger has a name and a collection of rules, like "Change status to: X" or "Play sound: Y".
- Each column may be bound to a trigger.
- Multiple columns may share the same trigger.
- Later UI refinements will make the cases around "copy trigger" vs "reference the same trigger" vs "create a new ad-hoc trigger" more clear.
- Triggers have their own edit policy.
- Triggers are always world-visible, like Herald rules.
Test Plan: Poked around, created some empty trigger objects, and nothing exploded. This doesn't actually do anything useful yet since triggers can't have any rule behavior and columns can't actually be bound to triggers.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T5474
Differential Revision: https://secure.phabricator.com/D20279
Summary: See downstream <https://phabricator.wikimedia.org/T166358>. The notifications menu is missing some CSS to color and style values in stories like "renamed task from X to Y".
Test Plan:
Before:
{F6302123}
After:
{F6302122}
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20310
Summary:
When a TOS-like Legalpad document is marked "Require this document to use Phabricator", the login prompt shows a "Manage" button, but that button doesn't work.
When we're presenting a document as a session gate, don't show "Manage".
Test Plan: Viewed a required document during a session gate (no "Manage" button) and normally (saw "Manage" button).
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20312
Summary:
See downstream <https://phabricator.wikimedia.org/T208254>.
I can't actually reproduce any issue here (we only show this field when creating a document, and only if the viewer is an administrator), so maybe this relied on some changes or was originally reported against older code.
Regardless, the validation isn't quite right: it requires administrator privileges to apply this transaction at all, but should only require administrator privileges to change the value.
Test Plan:
Edited Legalpad documents as an administrator and non-administrator before and after the change, with and without signatures being required.
Couldn't reproduce the original issue, but this version is generally more correct/robust.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20311
Summary: See downstream <https://phabricator.wikimedia.org/T182232>. We currently don't mark repository handles as closed.
Test Plan: Mentioned two repositories with `R1` (active) and `R2` (inactive). After patch, saw `R2` visually indicated as closed/inactive.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20309
Summary: See PHI1153. The "Runnable" and "Restartable" behaviors interact (to click "restart", you must be able to run the build AND it must be restartable). Make this more clear.
Test Plan: {F6301739}
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Differential Revision: https://secure.phabricator.com/D20307
Summary: Ref T13266. This callsite is using the older API; swap it to use pagers.
Test Plan: Viewed a Phame blog post with siblings, saw the previous/next posts linked.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: nicolast
Maniphest Tasks: T13263, T13266
Differential Revision: https://secure.phabricator.com/D20319
Summary:
Ref T13263. Two minor issues:
- The "reconcile" dialog shows the wrong sign because JS signs differ from normal signs (for example, PST or PDT or whatever we're in right now is shown as "UTC+7", but should be "UTC-7").
- The big dropdown of possible timezones lumps "UTC+X:30" timezones into "UTC+X".
Test Plan:
- Reconciled "America/Nome", saw negative UTC offsets for "America/Nome" and "America/Los_Angeles" (previously: improperly positive).
- Viewed the big timzone list, saw ":30" and ":45" timezones grouped/labeled more accurately.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13263
Differential Revision: https://secure.phabricator.com/D20314
Summary:
See <https://discourse.phabricator-community.org/t/error-when-sending-a-message-chat-room/2548>.
Conpherence calls `setAfterID()` and `setBeforeID()` directly on a subquery, but these methods no longer exist.
Use a pager instead. This code probably shouldn't exist (we should use some other approach to fetch this data in most cases) but that's a larger change.
Test Plan: Sent messages in a Conpherence thread. Before: fatal; after: success. Viewed the Conphrence menu, loaded threads, etc.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20318
Summary:
Ref T13091. The Ferret "rank" column is a function of the query text and looks something like `SELECT ..., 2 + 2 AS rank, ...`.
You can't apply conditions to this kind of dynamic column with a WHERE clause: you get a slightly unhelpful error like "column rank unknown in where clause". You must use HAVING:
```
mysql> SELECT 2 + 2 AS x WHERE x = 4;
ERROR 1054 (42S22): Unknown column 'x' in 'where clause'
mysql> SELECT 2 + 2 AS x HAVING x = 4;
+---+
| x |
+---+
| 4 |
+---+
1 row in set (0.00 sec)
```
Add a flag to paging column definitions to let them specify that they must be applied with HAVING, then apply the whole paging clause with HAVING if any column requires HAVING.
Test Plan:
- In Maniphest, ran a fulltext search matching more than 100 results, ordered by "Relevance", then clicked "Next Page".
- Before patch: query with `... WHERE rank > 123 OR ...` caused MySQL error because `rank` is not a WHERE-able column.
- After patch: query builds as `... HAVING rank > 123 OR ...`, pages properly, no MySQL error.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13091
Differential Revision: https://secure.phabricator.com/D20298
Summary:
See downstream <https://phabricator.wikimedia.org/T210482>.
On mobile, the task graph can take up most of the screen. Hide it on devices. Keep it on the standalone view if you're really dedicated and willing to rotate your phone or whatever to see the lines.
Test Plan: Dragged window real narrow, saw graph hide.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20313
Summary:
See <https://discourse.phabricator-community.org/t/unhandled-exception-on-show-older-changes/2545/>.
Before T13266, this query got away without having real paging because it used simple ID paging only and results are never actually hidden (today, you can always see all transactions on an object).
Provide `withIDs()` so the new, slightly stricter paging works.
Test Plan: On an object with "Show Older" in the transaction record, clicked the link. Before: exception in paging code (see Discourse link above). After: transactions loaded cleanly.
Reviewers: amckinley, avivey
Reviewed By: avivey
Differential Revision: https://secure.phabricator.com/D20317
Ref T13266. We never page these queries, and previously never reached the
"nextPage()" method. The call order changed recently and this method is now
reachable. For now, just no-op it rather than throwing.
After the cursor changes, we may fatal on pages with a large number of children
because "c.title" is not a selected column. We currently join the "content"
table if "updated" is part of the order vector, but not if "title" is part of
the order vector. This isn't right: "updated" is on the primary table, and only
"content" is on the joined table.
Summary:
Ref T13091. In Differential, if you provide a query and "Sort by: Relevance", we build a query like this:
```
((SELECT revision.* FROM ... ORDER BY rank) UNION ALL (SELECT revision.* FROM ... ORDER BY rank)) ORDER BY rank
```
The internal "ORDER BY rank" is technically redundant (probably?), but doesn't hurt anything, and makes construction easier.
The problem is that the outer "ORDER BY rank" at the end, which attempts to order the results of the two parts of the UNION, can't actually order them, since `rank` wasn't selected.
(The column isn't actually "rank", which //is// selected -- it's the document modified/created subcolumns, which are not.)
To fix this, actually select the fulltext columns into the result set.
Test Plan:
- Ran a non-empty fulltext query in Differential with "Bucket: Required Action" selected so the UNION construction fired.
- Ran normal queries in Maniphest and global search.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13091
Differential Revision: https://secure.phabricator.com/D20297
Summary:
Ref T13091. If you "Order By: Relevance" but don't actually specify a query, we currently raise a bare exception.
This operation is sort of silly/pointless, but it seems like it's probably best to just return the results for the other constraints in the fallback order (usually, by ID). Alternatively, we could raise a non-bare exception here ("You need to provide a fulltext query to order by relevance.")
Test Plan: Queried tasks by relevance with no actual query text.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13091
Differential Revision: https://secure.phabricator.com/D20296
Summary:
Ref T13259. Currently, visiting a page that executes a query with an invalid cursor raises a bare exception that escapes to top level.
Catch this a little sooner and tailor the page a bit.
Test Plan: Visited `/maniphest/?after=335234234223`, saw a nicer exception page.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13259
Differential Revision: https://secure.phabricator.com/D20295
Summary:
Ref T13259. Currently, queries set a flag and return a partial result set when they overheat. This is mostly okay:
- It's very unusual for queries to overheat if they don't have a real viewer.
- Overheating is rare in general.
- In most cases where queries can overheat, the context is a SearchEngine UI, which handles this properly.
In T13259, we hit a case where a query with an omnipotent viewer can overheat: if you have more than 1,000 consecutive commits in the database with invalid `repositoryID` values, we'll overheat and bail out. This is pretty bad, since we don't process everything.
Change this beahvior:
- Throw by default, so this stuff doesn't slip through the cracks.
- Handle the SearchEngine case explicitly ("it's okay to overheat, we'll handle it").
- Make `QueryIterator` disable overheating behavior: if we're iterating over all objects, we want to hit the whole table even if most of it is garbage.
There are some cases where this might cause new exception behavior that we don't necessarily want. For example, in Owners, each package shows "recent commits in this package". If you can't see the first 1,000 recent commits, you'd previously get a slow page with no results. Now you'll probably get an exception.
If these crop up, I think the best approach for now is to deal with them on a case-by-case basis and see how far we get. In the "Owners" case, it might be good to query by repositories you can see first, then query by commits in the package in those repositories. That should give us a better outcome than any generic behavior we could implement.
Test Plan:
- Added 100000 to all repositoryID values for commits on my local install.
- Before making changes, ran `bin/repository rebuild-identities --all --trace`. Saw the script process 1,000 rows and exit silently.
- Applied the first part ("throw by default") and ran `bin/repository rebuild-identities`. Saw the script process 1,000 rows, then raise an exception.
- Applied the second part ("disable for queryiterator") and ran the script again. Saw the script process all 15,000 rows without issues (although none are valid and none actually load).
- Viewed Diffusion, saw appropriate NUX / "overheated" UIs.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13259
Differential Revision: https://secure.phabricator.com/D20294
Summary:
Depends on D20292. Ref T13259. This converts the rest of the `getPagingValueMap()` callsites to operate on internal cursors instead.
These are pretty one-off for the most part, so I'll annotate them inline.
Test Plan:
- Grouped tasks by project, sorted by title, paged through them, saw consistent outcomes.
- Queried edges with "edge.search", paged through them using the "after" cursor.
- Poked around the other stuff without catching any brokenness.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13259
Differential Revision: https://secure.phabricator.com/D20293
Summary:
Depends on D20291. Ref T13259. Move all the simple cases (where paging depends only on the partial object and does not depend on keys) to a simple wrapper.
This leaves a smaller set of more complex cases where we care about external data or which keys were requested that I'll convert in followups.
Test Plan: Poked at things, but a lot of stuff is still broken until everything is converted.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13259
Differential Revision: https://secure.phabricator.com/D20292
Summary:
Ref T13259.
(NOTE) This is "infrastructure/guts" only and breaks some stuff in Query subclasses. I'll fix that stuff in a followup, it's just going to be a larger diff that's mostly mechanical.
When a user clicks "Next Page" on a tasks view and gets `?after=100`, we want to show them the next 100 //visible// tasks. It's possible that tasks 1-100 are visible, but tasks 101-788 are not, and the next visible task is 789.
We load task ID `100` first, to make sure they can actually see it: you aren't allowed to page based on objects you can't see. If we let you, you could use "order=title&after=100", plus creative retitling of tasks, to discover the title of task 100: create tasks named "A", "B", etc., and see which one is returned first "after" task 100. If it's "D", you know task 100 must start with "C".
Assume the user can see task 100. We run a query like `id > 100` to get the next 100 tasks.
However, it's possible that few (or none) of these tasks can be seen. If the next visible task is 789, none of the tasks in the next page of results will survive policy filtering.
So, for queries after the initial query, we need to be able to page based on tasks that the user can not see: we want to be able to issue `id > 100`, then `id > 200`, and so on, until we overheat or find a page of results (if 789-889 are visible, we'll make it there before overheating).
Currently, we do this in a not-so-great way:
- We pass the external cursor (`100`) directly to the subquery.
- We query for that object using `getPagingViewer()`, which is a piece of magic that returns the real viewer on the first page and the omnipotent viewer on the 2nd..nth page. This is very sketchy.
- The subquery builds paging values based on that object (`array('id' => 100)`).
- We turn the last result from the subquery back into an external cursor (`200`) and save it for the next time.
Note that the last step happens BEFORE policy (and other) filtering.
The problems with this are:
- The phantom-schrodinger's-omnipotent-viewer thing isn't explicity bad, but it's sketchy and generally not good. It feels like it could easily lead to a mistake or bug eventually.
- We issue an extra query each time we page results, to convert the external cursor back into a map (`100`, `200`, `300`, etc).
- In T13259, there's a new problem: this only works if the object is filtered out for policy reasons and the omnipotent viewer can still see it. It doesn't work if the object is filtered for some other reason.
To expand on the third point: in T13259, we hit a case where 100+ consecutive objects are broken (they point to a nonexistent `repositoryID`). These objects get filtered unconditionally. It doesn't matter if the viewer is omnipotent or not.
In that case: we set the next external cursor from the raw results (e.g., `200`). Then we try to load it (using the omnipotent viewer) to turn it into a map of values for paging. This fails because the object isn't loadable, even as the omnipotent viewer.
---
To fix this stuff, the new approach steps back a little bit. Primarily, I'm separating "external cursors" from "internal cursors".
An "External Cursor" is a string that we can pass in `?after=X` URIs. It generally identifies an object which the user can see.
An "Internal Cursor" is a raw result from `loadPage()`, i.e. before policy filtering. Usually, (but not always) this is a `LiskDAO` object that doesn't have anything attached yet and hasn't been policy filtered.
We now do this, broadly:
- Convert the external cursor to an internal cursor.
- Execute the query using internal cursors.
- If necessary, convert the last visible result back into an external cursor at the very end.
This fixes all the problems:
- Sketchy Omnipotent Viewer: We no longer ever use an omnipotent viewer. (We pick cursors out of the result set earlier, instead.)
- Too Many Queries: We only issue one query at the beginning, when going from "external" to "internal". This query is generally unavoidable since we need to make sure the viewer can see the object and that it's a real / legitimate object. We no longer have to query an extra time for each page.
- Total Failure on Invalid Objects: we now page directly with objects out of `loadPage()`, before any filtering, so we can page over invisible or invalid objects without issues.
This change switches us over to internal/external cursors, and makes simple cases (ID-based ordering) work correctly. It doesn't work for complex cases yet since subclasses don't know how to get paging values out of an internal cursor yet. I'll update those in a followup.
Test Plan: For now, poked around a bit. Some stuff is broken, but normal ID-based lists load correctly and page properly. See next diff for a more detailed test plan.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13259
Differential Revision: https://secure.phabricator.com/D20291
Summary:
See PHI1134. Previously, see T13082 and D19969 for some sort-of-related stuff.
Currently, edits work roughly like this:
- Suppose we're editing object X, and we're also going to edit some other object, Y, because X mentioned Y or the edit is making X a child or parent of Y, or unblocking Y.
- Do the actual edit to X, including inverse edits ("alice mentioned Y on X.", "alice added a child revision: X", etc) which apply to Y.
- Run Herald rules on X.
- Publish the edit to X.
The "inverse edits" currently do this whole process inline, in a sub-editor. So the flow expands like this:
- Begin editing X.
- Update properties on X.
- Begin inverse-edge editing Y.
- Update properties on Y.
- Run (actually, skip) Herald rules on Y.
- Publish edits to Y.
- Run Herald rules on X.
- Publish edits to X.
Notably, the "Y" stuff publishes before the "X" Herald rules run. This creates potential problems:
- Herald rules may change the name or visibility policy of "X", but we'll publish mail about it via the edits to Y before those edits apply. This is a problem only in theory, we don't ship any upstream rules like this today.
- Herald rules may "Require Secure Mail", but we won't know that at the time we're building mail about the indirect change to "Y". This is a problem in practice.
Instead, switch to this new flow, where we stop the sub-editors before they publish, then publish everything at the very end once all the edits are complete:
- Begin editing X.
- Update properties on X.
- Begin inverse-edge editing Y.
- Update properties on Y.
- Skip Herald on Y.
- Run Herald rules on X.
- Publish X.
- Publish all child-editors of X.
- Publish Y.
Test Plan:
- Created "Must Encrypt" Herald rules for Tasks and Revisions.
- Edited object "A", an object which the rules applied to directly, and set object "B" (a different object which the rules did not hit) as its parent/child and/or unblocked it.
- In `bin/mail list-outbound`, saw:
- Mail about object "A" all flagged as "Must Encrypt".
- Normal mail from object B not flagged "Must Encrypt".
- Mail from object B about changing relationships to object A flagged as "Must Encrypt".
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20283
