Summary:
Fixes T12087. When transitioning into a clustered configuration for the first time, the documentation recommends using a one-device cluster as a transitional step.
However, installs may not do this for whatever reason, and we aren't as clear as we could be in warning about clusterizing directly into a multi-device cluster.
Roughly, when you do this, we end up believing that working copies exist on several different devices, but have no information about which copy or copies are up to date. //Usually// they all were already synchronized and are all up to date, but we can't make this assumption safely without risking data.
Instead, we err on the side of caution, and require a human to tell us which copy we should consider to be up-to-date, using `bin/repository thaw --promote`.
Test Plan:
```
$ ./bin/repository clusterize rLOCKS --service repos001.phacility.net
Service "repos001.phacility.net" is actively bound to more than one device
(local002.local, local001.phacility.net).
If you clusterize a repository onto this service it will be unclear which
devices have up-to-date copies of the repository. This leader/follower
ambiguity will freeze the repository. You may need to manually promote a
device to unfreeze it. See "Ambiguous Leaders" in the documentation for
discussion.
Continue anyway? [y/N]
```
Read other changes.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12087
Differential Revision: https://secure.phabricator.com/D17169
Summary: Fixes T12090. In obscure situations lost to the mists of time, the `changes` column could be `null`. Force a string cast so the migration finishes, even though these changesets are likely meaningless.
Test Plan:
I did a force-reapply as a sanity check:
```
$ ./bin/storage upgrade -f --apply phabricator:20161213.diff.01.hunks.php
```
That went cleanly; it would only have caught dramatic errors, but I didn't completely butcher things.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12090
Differential Revision: https://secure.phabricator.com/D17168
Summary: Ref T5867. Rough in a Favorites application, not wired to anything.
Test Plan: tbd. currently 404s so... I messed up something. Tossing up to read.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T5867
Differential Revision: https://secure.phabricator.com/D17160
Summary: Fixes T12086. This got dropped by accident while cleaning up haunting.
Test Plan: Loaed a revision, hit "?", hit n/j/p/etc
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12086
Differential Revision: https://secure.phabricator.com/D17166
Summary:
See D17160. Previously, the `/edit/` route was never linked, but fataled when accessed. Make it 404 instead.
Also, fix an issue where editing "Application" menu items would fail because they didn't have a viewer.
Test Plan:
- Hit `/edit/`, got a 404.
- Edited an "Application" item.
- Moved, added, deleted, and edited other items.
Reviewers: chad
Reviewed By: chad
Differential Revision: https://secure.phabricator.com/D17165
Summary:
Fixes T10968. In rare situations, we can generate a diff, then hit an error which causes this update to fail.
When it does, we tend to get stuck in a loop creating diffs, which can fill the database up with garbage. We saw this once in the Phacility cluster, and one instance hit it, too.
Instead: when we create a diff, keep track of which commit we generated it from. The next time through, reuse it if we already built it.
Test Plan:
- Used `bin/differential attach-commit <commit> <revision>` to hit this code.
- Simulated a filesystem write failure, saw the diff get reused.
- Also did a normal update, which worked properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10968
Differential Revision: https://secure.phabricator.com/D17164
Summary:
Ref T12074. This uses more consistent Conduit keys for constraint names.
This is a minor compatibility break on watchers/members but since these methods are more useful now this is probably a good time to try to get away with it, and a more consistent API is better in the long run. I need to issue compatibility guidance for the milestones thing anyway and that one isn't avoidable, so try to rip the bandage off all in one go.
Test Plan: Reviewed new constraint names from console, called methods using them.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17161
Summary:
Fixes T12082. Ref T11114. When modular transaction render a handle list, they use HandleListView, which has a text mode.
However, the HandleListView is a TagView, and currently TagViews always render a tag of some kind. Allow them to return `null` to decline to render any tag.
Test Plan:
- Added a pile of debugging stuff to `ApplicationTransactionEditor` to throw during mail generation.
- Added a reviewer to a revision.
- Used `bin/worker execute --id ...` to hit the mail generation repeatedly.
- Before patch: mail generated with a <span>, even in text mode.
- After patch: clean mail generation.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12082, T11114
Differential Revision: https://secure.phabricator.com/D17162
Summary: Ref T11114. After evaluating typeahead tokens, we could process blocking reviewer removals incorrectly: we may get structures back.
Test Plan: Removed blocking reviewers from the web UI.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17163
Summary:
Ref T12074. The "v3" API methods (`*.search`, `*.edit`) are currently marked as "unstable", but they're pretty stable and essentially all new code should be using them.
Although these methods are seeing some changes, almost all changes are additive (support for new constraints or attachemnts) and do not break backward compatibility. We have no major, compatibility-breaking changes planned.
I don't want to mark the older methods "deprecated" yet since `arc` still uses a lot of them and there are some capabilities not yet available on the v3 methods, but introduce a new "frozen" status with pointers to the new methods.
Overall, this should gently push users toward the newer methods.
Test Plan: {F2325323}
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17158
Summary:
Ref T12074. Provide a basic but functional v3 API endpoint for reading workboard column information.
There is no equivalent to this in the UI yet, although there may be some day (perhaps adjacent to T5024).
Test Plan:
- Queried for all columns.
- Queried for columns on a particular board using `projectPHIDs`.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17157
Summary:
Ref T12074. This allows callers to identify which columns an object appears in (currently, always tasks).
There are a few major cases:
- Object is in a normal column: we return column information.
- Object is in a proxy column (subproject or milestone). For example, when you look at the board for "Some Parent Project", the task might show up in a milestone column. I've chosen to not return anything in this case: you can figure out that the task is there by looking at the project structure, and this is kind of an internal artifact of the implementation and probably not useful to callers.
- Project does not have a workboard: we return nothing.
These seem fairly reasonable, I think?
Test Plan:
- Queried for tasks, using the "columns" attachment.
- Dragged a task across a board, querying it repeatedly. Got expected results for normal column (the column), subprojects with no board (nothing), milestones with no board (nothing) and mielstones/subprojects with a board (the column on //that// board, only, not the proxy column on the parent).
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17156
Summary:
Ref T12074. Allows querying for project by direct parent (find only immediate children) or any ancestor (find all descendants) using the API.
There's no proper web UI for this since I'm not sure how useful it is, but you can `/project/?parent=PHID-PROJ-...` or `/project/?ancestor=...` for now. We can add UI later if/when use cases arise, but it's not immediately clear to me that this is useful to do from the web.
Test Plan:
- From API, queried with `parentPHIDs` and `ancestorPHIDs`, finding direct children only and all descendants, respectively.
- From web UI, fiddled with `?parent=...` and `?ancestor=...` to make sure they work too. This isn't intended to be a user-facing feature.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17155
Summary:
Ref T12074.
- Adds a new "parent" property on main results. This shows an abbreviated version of the project's parent, or `null` if the project is a root project.
- Adds a new "ancestor" attachment to pull the entire ancestor list.
- Adds a new "depth" property on main results.
- You can use "parent" or "depth" to tell if a project is a subproject or not.
These attempt to balance convenience, power, and performance: the full ancestor list can be big so I made it an attachment, but the other stuff isn't too big and is cheap and seems reasonable to always include.
Test Plan:
In API results:
- Saw null parent (root projects) and non-null parent (subprojects/milestones).
- Used "ancestors" attchment, got full list of ancestors.
- Saw appropriate "depth" values.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17154
Summary:
Ref T12074.
- `project.search` now returns milestones by default.
- A new constraint, `isMilestone`, allows filtering to milestones, non-milestones, or both (API and web UI).
- `project.search` now returns a milestone number for milestones, or `null` for non-milestones.
NOTE: Existing custom saved queries in projects which previously did not return milestones now will. I expect this to have little-to-no impact on users, and these queries are easy to correct, but I'll note this in changelogs.
Test Plan:
- Ran various queries with `project.search` and in the web UI, searching for milestones, non-milestones, and both.
- Web UI default behavior (no milestones) is unchanged, but you can now get milestones if you want them.
- Queried a milestone by ID/PHID via API.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12074
Differential Revision: https://secure.phabricator.com/D17153
Summary:
Ref T9058. The stricter filtering is over-filtering Handles. For example, in the Phacility cluster, users can not see Almanac services.
So this filtering happens:
- The AlmanacServiceQuery filters the service beacuse they can't see the application.
- The HandleQuery generates a "you can't see this" handle.
- But then the HandleQuery filters that handle! It has a "service" PHID and the user can't see Almanac.
This violates the assumption that all application code makes about handles: it's OK to query handles for objects you can't see, and you'll get something back.
Instead, don't do application filtering on handles.
Test Plan:
- Added a failing test and made it pass.
- As a user who can not see Almanac, viewed an Instances timeline.
- Before patch: fatal on trying to load a handle for a Service.
- After patch: smooth sailing.
Reviewers: chad
Maniphest Tasks: T9058
Differential Revision: https://secure.phabricator.com/D17152
Summary: Fixes T12080. This was missing a "/", but stop hard-coding these URIs.
Test Plan: Clicked both links with Quickling as a logged-in and logged-out user, ended up in the right place.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12080
Differential Revision: https://secure.phabricator.com/D17151
Summary: Ref T5867, adds a customPHID field, nullable, and lets you query by it... i think? Not fully able to grok all the EditEngine stuff, but I think this is the right place for the query.
Test Plan: Not wired to anything, but pulling up project menu, editing, all still works.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T5867
Differential Revision: https://secure.phabricator.com/D17149
Summary: Ref T11816. The underlying format of recurrence end dates swapped around a bit and we now try to compare `null` to a valid date if you're setting it for the first time.
Test Plan:
- On a new event, set a recurrence end date.
- Then, removed a recurrence end date.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11816
Differential Revision: https://secure.phabricator.com/D17150
Summary:
Fixes T10687. Fixes T12064.
- Primarily, adds a margin around the edge of the screen for the purposes of aligning the tooltip.
- Also, tries to flip the tooltip if it can (e.g., if the tooltip normally goes east, try west first), then tries other positions exhastively.
Test Plan:
{F2309363}
{F2309364}
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12064, T10687
Differential Revision: https://secure.phabricator.com/D17145
Summary:
Ref T11114. When you comment, we try to upgrade your review status to "commented".
This can conflict with upgrading it to "accepted" or "rejected", or removing it entirely.
For now, just avoid making this update. After T10967, I expect "you commented" to be orthogonal to accepted/rejected so it should stop conflicting on its own.
Test Plan:
- As an "added" reviewer, accepted a revision with a comment in the same transaction.
- Before patch: accept didn't stick.
- After patch: accept sticks.
This may be somewhat magical/order-dependent but I was able to reproduce it locally.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17146
Summary: Fixes T12068. These are inbound messages, not outbound.
Test Plan: Read carefully.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12068
Differential Revision: https://secure.phabricator.com/D17144
Summary: Fixes T12065. Some of the normal behavior of these actions got juggled when I made "Accept" and "Reject" conflict.
Test Plan:
- Added "Accept".
- Added "Reject", saw it remove "Accept".
- Added "Change Projects".
- Added "Assign/Claim".
- Removed "Assign/Claim".
- Removed "Accept".
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12065
Differential Revision: https://secure.phabricator.com/D17143
Summary: Minor color saturation here, ideal for low quality monitors.
Test Plan:
Review new colors in various scenarios.
{F2305178}
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Differential Revision: https://secure.phabricator.com/D17141
Summary:
Fixes T10136. This reinforces ongoing or failed builds in the comment action area.
We already emit a similar message for unit test failures from `arc unit`. This should probably obsolete that, eventually.
Test Plan:
{F2304809}
{F2304810}
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10136
Differential Revision: https://secure.phabricator.com/D17140
Summary: When a user queries by package monogram explicitly, search by package ID.
Test Plan: {F2305075}
Reviewers: chad
Reviewed By: chad
Differential Revision: https://secure.phabricator.com/D17142
Summary:
Fixes T9276. Fixes T8650. The story so far:
- We once published build updates to Revisions.
- An unrelated fix (D10911) sent them to the Diffs instead of Revisions, which isn't useful, since you can't see a diff's timeline anywhere.
- This also caused a race condition, where the RevisionEditor and DiffEditor would update the diff simultaneously (T8650).
- The diff update was just disabled to avoid the race (part of D13441).
- Instead, allow the updates to go somewhere else. In this case, we send commit updates to the commit but send diff updates to the revision so you can see 'em.
- Since everything will be using the revision editor now, we should either get proper lock behavior for free or it should be easy to add if something whack is still happening.
- Overall, this should pretty much put us back in working order like we were before D10911.
This behavior is undoubtedly refinable, but this should let us move forward.
Test Plan:
Saw a build failure in timeline:
{F2304575}
Reviewers: chad
Reviewed By: chad
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T9276, T8650
Differential Revision: https://secure.phabricator.com/D17139
Summary: Adds a FormEditEngine MenuItem for adding forms to Projects, Home, QuickCreate. Also adds an EditEngine typeahead that has token rendering issues currently.
Test Plan: Set a normal form as a menu item, edit it, set the name. Set a custom form as a menu item, edit it, set a name.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Differential Revision: https://secure.phabricator.com/D17098
Summary: Tweaks the diff colors here a bit, as well as making full diffs slightly easier to read in full. Ref T12060
Test Plan:
Tested prose diffs, email prose diffs, and a regular Differential revision.
{F2304056}
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T12060
Differential Revision: https://secure.phabricator.com/D17138
Summary:
Fixes T12062. Like the commits from the year 3500, you can artificially build commits with no date information.
We could explicitly store these as `null` to fully respect the underlying datastore. However, I think it's very unlikely that these commits are intentional/meaningful or that this is valuable.
Additionally, "git show" interprets these commits as "Jan 1, 1970". Just store a `0` to mimic its behavior.
Test Plan:
- Following the process in T11537#192019, artificially created a commit with //no// date information (I deleted all date information from the message).
- Used `git show` / `git log --format ...` to inspect it: "Jan 1, 1970" on `git show`, no information at all on `%aD`, `%aT`, etc.
- Pushed it.
- Saw exception for trying to insert empty string into epoch colum from `bin/repository update`.
- Applied patch.
- Got a clean import.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12062
Differential Revision: https://secure.phabricator.com/D17136
Summary:
Fixes T12058. When the user visits `/maniphest`, for example, we redirect to `/maniphest/`.
Since this redirect is very low-level (at the Aphront level, below the Site level) we need to preserve the request Host rather than correct it to `PhabricatorEnv::getURI()` or similar -- the request may be hiting a different Site like a blog domain.
Currently, we do not preserve the port. Instead, preserve the port if it is not a standard port for the protocol (80 for http, 443 for https).
Test Plan:
- Made a request with a missing slash and a normal port in my browser, got redirected normally.
- Made a request with a missing slash and a nonstandard port, got redirected on the same port.
```
$ curl -H 'Host: local.phacility.com:123' -v http://local.phacility.com/diffusion
* Trying 127.0.0.1...
* Connected to local.phacility.com (127.0.0.1) port 80 (#0)
> GET /diffusion HTTP/1.1
...
>
< HTTP/1.1 302 Found
...
< Location: http://local.phacility.com:123/diffusion/
...
```
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12058
Differential Revision: https://secure.phabricator.com/D17134
Summary: Ref T3612. Doesn't render correctly, need help please. Adds a download icon into the renderfilelinkview to allow easier downloads.
Test Plan: Click on link, get download, click on file, get lightbox.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T3612
Differential Revision: https://secure.phabricator.com/D16980
Summary: Ref T11114. Move email/command actions, like "!reject", to modular transactions + editengine.
Test Plan: Used `bin/mail receive-test` to pipe "!stuff" to an object, saw appropraite effects in web UI.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17133
Summary:
Ref T11114. When a user selects "Accept", and then selects "Reject", remove the "Accept". It does not make sense to both accept and reject a revision.
For now, every one of the "actions" conflicts: accept, reject, resign, claim, close, commandeer, etc, etc. I couldn't come up with any combinations that it seems like users are reasonably likely to want to try, and we haven't received combo-action requests in the past that I can recall.
Test Plan:
- Selected "Accept", then selected "Reject". One replaced the other.
- Selected "Accept", then selected "Change Subscribers". Both co-existed happily.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17132
Summary:
Ref T11114. Recent changes broke the links to jump to inline comments from the previews because they get hooked up with JS.
Restore the linking behavior.
Test Plan: Clicked "View" on an inline comment preview, jumped to that comment.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17131
Summary:
Fixes T9058. Normally, "Query" classes apply an application check and just don't load anything if it fails.
However, in some cases (like email recipient filtering) we run policy checks without having run a Query check first. In that case, one user (the actor) loads the object, then we filter it against other users (the recipeints).
Explicitly apply the application check during normal filtering.
Test Plan: Added a failing test case and made it pass.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T9058
Differential Revision: https://secure.phabricator.com/D17127
Summary:
Fixes T11344. In the web UI, if a field like "Subscribers" on an object (like a task) contains values you don't have permission to see, you see tokens for them (like "Restricted Project") but not their names.
Make commit messages work the same way: you see the PHID, and can remove it or leave it there, but can't see the underlying name.
(We have to render an actual PHID rather than just "Restricted Thing" because we have to be able to figure out what edit the user is actually trying to make.)
Test Plan: Interacted with a revision via the CLI that had project reviewers I couldn't see.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11344
Differential Revision: https://secure.phabricator.com/D17124
Summary:
Fixes T9648. Diffs currently use `return $this->getRevision()->getViewPolicy();` to inherit their revision's view policy.
After the introduction of object policies, this is wrong for policies like "Subscribers", because it means "Subscribers to this object, the diff". Since Diffs have no subscribers, this always fails.
Instead, use extended policies so that the object policy evaluates in the context of the correct object (the revision).
Test Plan:
- Create a revision.
- Subscribe `alice` to it.
- Set view policy to "Subscribers".
- View revision as `alice`.
- Before patch: nonsense fatal about missing diff because of policy error.
- After patch: `alice` can see the revision.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T9648
Differential Revision: https://secure.phabricator.com/D17123
Summary: Fixes T10312. If your first line is "Reviewers: xyz", it's a title, not a "Reviewers" field.
Test Plan: Added unit test.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10312
Differential Revision: https://secure.phabricator.com/D17122
Summary:
Fixes T8360. We will now parse revisions out of "Differential Revision: X" followed by other ad-hoc fields which we do not recognize. Previously, these fields would be treated as part of the value.
(In the general case, other fields may line wrap so we can't assume that fields are only one line long. However, we can make that assumption safely for this field.)
Also maybe fix whatever was going on in T9965 although that didn't really have a reproduction case.
Test Plan: Added unit tests.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T8360
Differential Revision: https://secure.phabricator.com/D17121
Summary: Ref T11114. Fixes T10323.
Test Plan:
- Marked comments as done only: no warning about not leaving a comment.
- Did nothing: warning about posting an empty comment.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114, T10323
Differential Revision: https://secure.phabricator.com/D17120
Summary: Ref T11114. Although I plan to rewrite this system eventually (T10448) it's easy enough to punt for now.
Test Plan: punt
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17119
Summary:
Ref T11114. This restores:
- Commandeering should exeucte Herald.
- Commandeering should swap reviewers.
- "Request Review" on an "Accepted" revision should downgrade reviewers so they have to accept again.
Test Plan:
- Commandeered, saw Herald run and reviewers swap.
- Requested review of an accepted revision, saw it drop down to "Needs Review" with "Accepted Prior" on the reviewer.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17118
