Summary:
Depends on D20040. Ref T13242. See PHI1039. See PHI873. Two reasonable cases have arisen recently where extending validation rules would help solve problems.
We can do this in a pretty straightforward way with a standard extension pattern.
Test Plan:
Used this extension to keep ducks away from projects:
```lang=php
<?php
final class NoDucksEditorExtension
extends PhabricatorEditorExtension {
const EXTENSIONKEY = 'no.ducks';
public function getExtensionName() {
return pht('No Ducks!');
}
public function supportsObject(
PhabricatorApplicationTransactionEditor $editor,
PhabricatorApplicationTransactionInterface $object) {
return ($object instanceof PhabricatorProject);
}
public function validateTransactions($object, array $xactions) {
$errors = array();
$name_type = PhabricatorProjectNameTransaction::TRANSACTIONTYPE;
$old_value = $object->getName();
foreach ($xactions as $xaction) {
if ($xaction->getTransactionType() !== $name_type) {
continue;
}
$new_value = $xaction->getNewValue();
if ($old_value === $new_value) {
continue;
}
if (preg_match('/duck/i', $new_value)) {
$errors[] = $this->newInvalidTransactionError(
$xaction,
pht('Project names must not contain the substring "duck".'));
}
}
return $errors;
}
}
```
{F6162585}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13242
Differential Revision: https://secure.phabricator.com/D20041
Summary:
Depends on D20035. Ref T13222.
- Allow individual transactions to request one-shot MFA if available.
- Make "change username" request MFA.
Test Plan:
- Renamed a user, got prompted for MFA, provided it.
- Saw that I no longer remain in high-security after performing the edit.
- Grepped for other uses of `PhabricatorUserUsernameTransaction`, found none.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D20036
Summary:
Depends on D20023. Ref T13222. Although I think this isn't strictly necessary from a pure security perspective (since you can't modify the primary number while you have MFA SMS), it seems like a generally good idea.
This adds a slightly new MFA mode, where we want MFA if it's available but don't strictly require it.
Test Plan: Disabled, enabled, primaried, unprimaried, and edited contact numbers. With MFA enabled, got prompted for MFA. With no MFA, no prompts.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D20024
Summary:
Fixes T13082. When you create a revision (say, `D111`) with `Ref T222` in the body, we write a `D111 -> T222` edge ("revision 111 references task 222") and an inverse `T222 -> D111` edge ("task 222 is referenced by revision 111").
We also apply a transaction to `D111` ("alice added a task: Txxx.") and an inverse transaction to `T222` ("alice added a revision: Dxxx").
Currently, it appears that the inverse transaction can sometimes generate mail faster than `D111` actually commits its (database) transactions, so the mail says "alice added a revision: Unknown Object (Differential Revision)". See T13082 for evidence that this is true, and a reproduction case.
To fix this, apply the inverse transaction (to `T222`) after we commit the main object (here, `D111`).
This is tricky because when we apply transactions, the transaction editor automatically "fixes" them to be consistent with the database state. For example, if a task already has title "XYZ" and you set the title to "XYZ" (same title), we just no-op the transaction.
It also fixes edge edits. The old sequence was:
- Open (database) transaction.
- Apply our transaction ("alice added a task").
- Apply the inverse transaction ("alice added a revision").
- Write the edges to the database.
- Commit (database) transaction.
Under this sequence, the inverse transaction was "correct" and didn't need to be fixed, so the fixing step didn't touch it.
The new sequence is:
- Open (database) transaction.
- Apply our transaction ("alice added a task").
- Write the edges.
- Commit (database) transaction.
- Apply the inverse transaction ("alice added a revision").
Since the inverse transaction now happens after the database edge write, the fixing step detects that it's a no-op and throws it away if we do this naively.
Instead, add some special cases around inverse edits to skip the correction/fixing logic, and just pass the "right" values in the first place.
Test Plan:
Added and removed related tasks from revisions, saw appropriate transactions render on both objects.
(It's hard to be certain this completely fixes the issue since it only happened occasionally in the first place, but we can see if it happens any more on `secure`.)
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13082, T222
Differential Revision: https://secure.phabricator.com/D19969
Summary:
Fixes T12921. Currently, we call `getTitleForHTMLMail()`, but that calls `getTitleForMail()` which forces us into text rendering mode.
Instead, have `getTitleForHTML/TextMail()` force the rendering mode, then call `getTitleForMail()` with the desired rendering mode.
This causes stories like "epriestely added dependent tasks: x, y." to appear as links in email instead of plain text.
Test Plan: Used `bin/mail show-outbound --id ... --dump-html > out.html` to verify HTML mail.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T12921
Differential Revision: https://secure.phabricator.com/D19968
Summary:
See T1024. When "CAN_EDIT" became default in T13186, this was missed as an exception.
Watching shouldn't require "CAN_EDIT", so exempt it.
Test Plan:
- Before change: tried to watch a project I could not edit, got a policy error.
- After change: watched/unwatched a project I could not edit.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D19977
Summary:
Depends on D19908. Ref T13222. In D19897, I reordered some transaction code and affected the call order of `willApplyTransactions()`.
It turns out that we use this call for only one thing, and that thing is pretty silly: naming the raw paste data file when editing paste content.
This is only user-visible in the URL when you click "View Raw Paste" and seems exceptionally low-value, so remove the hook and pick a consistent name for the paste datafiles. (We could retain the name behavior in other ways, but it hardly seems worthwhile.)
Test Plan:
- Created and edited a paste.
- Grepped for `willApplyTransactions()`.
Note that `EditEngine` (vs `ApplicationTransacitonEditor`) still has a `willApplyTransactions()`, which has one callsite in Phabricator (in Calendar) and a couple in Instances. That's untouched and still works.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19909
Summary:
Depends on D19900. Ref T13222. See PHI873. When an object requires MFA, we currently require MFA for every transaction.
This includes some ambiguous cases like "unsubscribe", but also includes "mention", which seems like clearly bad behavior.
Allow an "MFA" object to be the target of mentions, "edit child tasks", etc.
Test Plan:
- Mentioned an MFA object elsewhere (no MFA prompt).
- Made an MFA object a subtask of a non-MFA object (no MFA prompt).
- Tried to edit an MFA object normally (still got an MFA prompt).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19901
Summary:
Depends on D19898. Ref T13222. See PHI873. Allow objects to opt into an "MFA is required for all edits" mode.
Put tasks in this mode if they're in a status that specifies it is an `mfa` status.
This is still a little rough for now:
- There's no UI hint that you'll have to MFA. I'll likely add some hinting in a followup.
- All edits currently require MFA, even subscribe/unsubscribe. We could maybe relax this if it's an issue.
Test Plan:
- Edited an MFA-required object via comments, edit forms, and most/all of the extensions. These prompted for MFA, then worked correctly.
- Tried to edit via Conduit, failed with a reasonably comprehensible error.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19899
Summary:
Depends on D19896. Ref T13222. See PHI873. Add a core "Sign With MFA" transaction type which prompts you for MFA and marks your transactions as MFA'd.
This is a one-shot gate and does not keep you in MFA.
Test Plan:
- Used "Sign with MFA", got prompted for MFA, answered MFA, saw transactions apply with MFA metadata and markers.
- Tried to sign alone, got appropriate errors.
- Tried to sign no-op changes, got appropriate errors.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19897
Summary:
Depends on D19919. Ref T11351. This method appeared in D8802 (note that "get...Object" was renamed to "get...Transaction" there, so this method was actually "new" even though a method of the same name had existed before).
The goal at the time was to let Harbormaster post build results to Diffs and have them end up on Revisions, but this eventually got a better implementation (see below) where the Harbormaster-specific code can just specify a "publishable object" where build results should go.
The new `get...Object` semantics ultimately broke some stuff, and the actual implementation in Differential was removed in D10911, so this method hasn't really served a purpose since December 2014. I think that broke the Harbormaster thing by accident and we just lived with it for a bit, then Harbormaster got some more work and D17139 introduced "publishable" objects which was a better approach. This was later refined by D19281.
So: the original problem (sending build results to the right place) has a good solution now, this method hasn't done anything for 4 years, and it was probably a bad idea in the first place since it's pretty weird/surprising/fragile.
Note that `Comment` objects still have an unrelated method with the same name. In that case, the method ties the `Comment` storage object to the related `Transaction` storage object.
Test Plan: Grepped for `getApplicationTransactionObject`, verified that all remaining callsites are related to `Comment` objects.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T11351
Differential Revision: https://secure.phabricator.com/D19920
Summary:
Depends on D19861. Ref T13222. See PHI996. Fixes T10743. Currently, notifications only work if a story also has a feed rendering.
Separate "visible in feed" and "visible in notifications", and make notifications query only notifications and vice versa.
Then, set the test notification stories to be visible in notifications only, not feed.
This could be refined a bit (there's no way to have the two views render different values today, for example) but since the only actual use case we have right now is test notifications I don't want to go //too// crazy future-proofing it. I could imagine doing some more of this kind of stuff in Conpherence eventually, though, perhaps.
Test Plan: Sent myself test notifications, saw them appear on my profile timeline and in the JS popup, and in my notifications menu, but not in feed.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222, T10743
Differential Revision: https://secure.phabricator.com/D19864
Summary:
Depends on D19860. Ref T13222. Ref T10743. See PHI996.
Long ago, there were different types of feed stories. Over time, there was less and less need for this, and nowadays basically everything is a "transaction" feed story. Each story renders differently, but they're fundamentally all about transactions.
The Notification test controller still uses a custom type of feed story to send notifications. Move away from this, and apply a transaction against the user instead. This has the same ultimate effect, but involves less weird custom code from ages long forgotten.
This doesn't fix the actual problem with these things showing up in feed. Currently, stories always use the same rendering for feed and notifications, and there need to be some additional changes to fix this. So no behavioral change yet, just slightly more reasonable code.
Test Plan: Clicked the button and got some test notifications, with Aphlict running.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222, T10743
Differential Revision: https://secure.phabricator.com/D19861
Summary:
Depends on D19858. Ref T13222. See PHI995. In D19635 and related revisions, inline behavior changed to allow you to pre-mark your own inlines as done (as a reviewer) and to pre-mark your inlines for you (as an author).
These actions generate low-value stories in the timeline, like "alice marked 3 comments done." when an author adds some notes to their own revision. These aren't helpful and can be a little misleading.
Instead, just don't count it when someone marks their own inlines as "done". If we throw away all the marks after throwing away the self-marks, hide the whole story.
This happens in three cases:
# You comment on your own revision, and don't uncheck the "Done" checkbox.
# You comment on someone else's revision, and check the "Done" checkbox before submitting.
# You leave a not-"Done" inline on your own revision, then "Done" it later.
Cases (1) and (2) seem unambiguously good/clear. Case (3) is a little more questionable, but I think this still isn't very useful for reviewers.
If there's still a clarity issue around case (3), we could change the story text to "alice marked 3 inline comments by other users as done.", but I think this is probably needlessly verbose and that no one will be confused by the behavior as written here.
(Also note that this story is never shown in feed.)
Test Plan: Created and marked a bunch of inlines as "Done" in Differential and Diffusion, as the author and reviewer/auditor. My own marks didn't generate timeline stories; marking others' comments still does.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19859
Summary:
Ref T13222. See PHI995. Before making a change to inline rendering, consolidate this code for generating the "alice added inlines comments." and "alice marked X inlines as done." transactions.
Both Differential and Diffusion have four very similar chunks of code. Merge them into shared methods and reduce code duplication across the methods.
(In the next change, I plan to hide the "done" story when the mark affects your own inline, since users marking their own inlines as "done" is generally not very interesting or useful.)
Test Plan: As author and reviewer/auditor, added inlines in Differential and Diffusion. As author, marked own and others inlines as done and undone. Got sensible transaction rendering and persistence of "Done".
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222
Differential Revision: https://secure.phabricator.com/D19858
Summary: Ref T13222. Ref T12588. See PHI683. To make "Create Subtask..." fancier, we need slightly more logic around subtype maps. Upgrade the plain old array into a proper object so it can have relevant methods, notably "get a list of valid child subtypes for some parent subtype".
Test Plan: Created and edited tasks, changed task subtypes. Grepped for affected symbols (`newEditEngineSubtypeMap`, `newSubtypeMap`).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13222, T12588
Differential Revision: https://secure.phabricator.com/D19852
Summary:
Ref T13216. See PHI970. Ref T13054. See some discussion in T13216.
When a Harbormaster Buildable object is first created for a Diff, it has no `containerPHID` since the revision has not yet been created.
We later (after creating a revision) send the Buildable a message telling it that we've added a container and it should re-link the container object.
Currently, we send this message in `applyExternalEffects()`, which runs inside the Differential transaction. If Harbormaster races quickly enough, it can read the `Diff` object before the transaction commits, and not see the container update.
Add a `didCommitTransaction()` callback after the transactions commit, then move the message code there instead.
Test Plan:
- See T13216 for substantial evidence that this change is on the right track.
- Before change: added `sleep(15)`, reproduced the issue reliably.
- After change: unable to reproduce issue even with `sleep(15)` (the `containerPHID` always populates correctly).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13216, T13054
Differential Revision: https://secure.phabricator.com/D19807
Summary:
Ref T13197. See PHI873. Record when a user has MFA'd and add a little icon to the transaction, similar to the exiting "Silent" icon.
For now, this just makes this stuff more auditable. Future changes may add ways to require MFA for certain specific transactions, outside of the ones that already always require MFA (like revealing credentials).
Test Plan: {F5877960}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13197
Differential Revision: https://secure.phabricator.com/D19665
Summary:
Depends on D19585. Ref T13164.
Almost all transactions require CAN_EDIT on the object, but they generally do not enforce this directly today. Instead, this is effectively enforced by Controllers, API methods, and EditEngine doing a `CAN_EDIT` check when loading the object to be edited.
A small number of transactions do not require CAN_EDIT, and instead require only a weaker/lesser permission. These are:
- Joining a project which you have CAN_JOIN on.
- Leaving a project which isn't locked.
- Joining a Conpherence thread you can see (today, no separate CAN_JOIN permission for Conpherence).
- Leaving a Conpherence thread.
- Unsubscribing.
- Using the special `!history` command from email.
Additionally, these require CAN_INTERACT, which is weaker than CAN_EDIT:
- Adding comments.
- Subscribing.
- Awarding tokens.
Soon, I want to add "disabling users" to this list, so that you can disable users if you have "Can Disable User" permission, even if you can not otherwise edit users.
It's possible this list isn't exhaustive, so this change might break something by adding a policy check to a place where we previously didn't have one. If so, we can go weaken that policy check to the appropriate level.
Enforcement of these special cases is currently weird:
- We mostly don't actually enforce CAN_EDIT in the Editor; instead, it's enforced before you get to the editor (in EditEngine/Controllers).
- To apply a weaker requirement (like leaving comments or leaving a project), we let you get through the Controller without CAN_EDIT, then apply the weaker policy check in the Editor.
- Some transactions apply a confusing/redundant explicit CAN_EDIT policy check. These mostly got cleaned up in previous changes.
Instead, the new world order is:
- Every transaction has capability/policy requirements.
- The default is CAN_EDIT, but transactions can weaken this explicitly they want.
- So now we'll get requirements right in the Editor, even if Controllers or API endpoints make a mistake.
- And you don't have to copy/paste a bunch of code to say "yes, every transaction should require CAN_EDIT".
Test Plan:
- Tried to add members to a Conpherence thread I could not edit (permissions error).
- Left a Conpherence thread I could not edit (worked properly).
- Joined a thread I could see but could not edit (worked properly).
- Tried to join a thread I could not see (permissions error).
- Implemented `requireCapabilites()` on ManiphestTransactionEditor and tried to edit a task (upgrade guidance error).
- Mentioned an object I can not edit on another object (works).
- Mentioned another object on an object I can not edit (works).
- Added a `{F...}` reference to an object I can not edit (works).
- Awarded tokens to an object I can not edit (works).
- Subscribed/unsubscribed from an object I can not edit (works).
- Muted/unmuted an object I can not edit (works).
- Tried to do other types of edits to an object I can not edit (correctly results in a permissions error).
- Joined and left a project I can not edit (works).
- Tried to edit and add members to a project I can not edit (permissions error).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13164
Differential Revision: https://secure.phabricator.com/D19586
Summary:
See PHI751. Ref T13164. We added a "silent" flag for Editors somewhat recently (currently reachable only for bulk edits with `bin/bulk ...` command).
However, this flag doesn't carry through to the sub-editor when we make inverse edge edits. These are edits like "X is a parent of Y", which cause an implicit "Y is a child of X" edit to occur.
Pass the flag through.
Test Plan:
- Rigged the relationships controller to make silent edits.
- Changed the parents of a revision from the web UI. Saw no mail or feed stories.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13164
Differential Revision: https://secure.phabricator.com/D19543
Summary:
Depends on D19372. Ref T13124. See PHI505. Currently, if you `!history` a task with a lot of comments, you get output like this:
> alice added a comment.
> bailey added a comment.
> alice added a comment.
> alice added a comment.
>
> AAAA
>
> BBBB
>
> AAAA
>
> AAAA
This is impossible to read. Put the "alice added a comment." headers above the actual comments for comments after the first.
These types of mail messages are unusual, but occur in several cases:
- The new `!history` command.
- Multiple comments on a draft revision before it promotes out of draft.
- (Probably?) Conduit API updates which submit multiple comment transactions for some reason.
Test Plan: Used `bin/mail receive-test` to send a `!history` command to a task, saw a much more readable rendering of the transaction log in the resulting email.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13124
Differential Revision: https://secure.phabricator.com/D19373
Summary:
See PHI505. Ref T13124. If you're an agent of a hostile state trying to exfiltrate corporate secrets, you might find yourself foiled if Phabricator is secured behind a VPN.
To assist users in this situation, provide a "!history" command which will dump the entire history of an object in a nice text format and get through the troublesome VPN.
Some issues with this:
- You currently get all the "X added a comment." up top, and then all the comments below. This isn't terribly useful.
- This goes through the "Must Encrypt" flag, but possibly should not? (On the other hand, this is a pretty willful way to bypass it the flag.)
Test Plan: Used `bin/mail receive-test ...` to send `!history` commands, got somewhat-useful response mail.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13124
Differential Revision: https://secure.phabricator.com/D19372
Summary:
Depends on D19279. Ref T13110. This implements the existing publishing logic for buildables, but does so via ModularTransactions instead of a core transaction type.
Since each application is implementing build transactions independently, this removes the core type.
Next, Differential will get a similar treatment.
Test Plan: Used `bin/harbormaster publish` (with some commenting-out-guard-clauses) to publish a commit Buildable; saw unchanged feed behavior.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13110
Differential Revision: https://secure.phabricator.com/D19280
Summary: Ref T13114. See PHI510. Firing Herald on mentioned objects tends to feel arbitrary and can substantially slow down edits which mention many objects.
Test Plan: Mentioned tasks on other tasks; verified that the normal path is hit normally, the new Herald-free path is hit on the mentioned object, and both still work fine and show up in the timeline.
Maniphest Tasks: T13114
Differential Revision: https://secure.phabricator.com/D19263
Summary:
See PHI466. Ref T13108. Somewhat recently, new rules were added so that "Resigning" from a revision takes you off the default recipient list, even if you're still a member of a project or package that is still a reviewer or subscriber.
However, these rules don't currently apply to the similar expansion which occurs in notifications. If you resign from a revision you may still get some notifications (just not email) if a package or project you're a member of is a reviewer or subscriber.
(Possibly these should eventually share more code, but just get things working for now.)
Test Plan:
- Created a revision as A.
- Added B as a reviewer.
- Added a package B is an owner for as a reviewer.
- As B, resigned. (Make sure B is also not an explicit subscriber.)
- Commented on the revision as A.
- Before: B is included in the expanded notification recipient list.
- After: B is no longer included in the expanded notification recipient list.
Maniphest Tasks: T13108
Differential Revision: https://secure.phabricator.com/D19244
Summary: See PHI433. Ref T13102. Users in the wild have mixed expecations about exactly what "draft" means. Recent changes have tried to make behavior more clear. As part of clarifying messaging, make it explicit that `@mention` does not work on drafts by showing users a warning when they try to `@mention` a user.
Test Plan:
- Mentioned users on drafts, got a warning.
- Posted normal comments on drafts, no warning.
- Posted normal/mention comments on non-drafts, no warning.
Maniphest Tasks: T13102
Differential Revision: https://secure.phabricator.com/D19210
Summary:
Fixes T10969. Ref T13077. When you create a Phriction document with a relative link (`[[ ./path/to/page ]]`) the initial email currently points to the wrong place.
This is because the context object (the page) isn't passed to the markup engine. Without this context, the relative link is rendered as though it appeared somewhere else (like a task or revision) where relative links don't make sense.
Test Plan: Created a new Phriction document with a relative link to `[[ ./porcupine_facts/starmap ]]`, saw a usable link in the resulting email.
Maniphest Tasks: T13077, T10969
Differential Revision: https://secure.phabricator.com/D19105
Summary:
Depends on D19064. Ref T13054. See that task for additional discussion.
When buildables are created by `arc` and have lint/unit messages, they can currently pass or fail before Herald triggers actual builds. This puts them in a pre-build state where they can't complete until Herald says it's okay.
On its own, this change intentionally strands `arc diff --only` diffs in the "PREPARING" stage forever.
Test Plan:
- Ran a build with `bin/harbormaster`, saw it build normally.
- Ran a build with web UI, saw it build normally.
- Ran a build with `arc diff`, saw it build normally.
- Ran a build with `arc diff --only`, saw it hang in "PREPARING" forever.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13054
Differential Revision: https://secure.phabricator.com/D19065
Summary: Ref T13054. Fixes T12714. Applies read locks to all transactions instead of only a very select subset (chat messages in Conpherence).
Test Plan: See <T13054#235650> for discussion and testing.
Maniphest Tasks: T13054, T12714
Differential Revision: https://secure.phabricator.com/D19059
Summary: Depends on D19049. Ref T11330. Adds some documentation for webhooks.
Test Plan: Read the documentation and found it to be exceptionally accurate and helpful.
Maniphest Tasks: T11330
Differential Revision: https://secure.phabricator.com/D19050
Summary: Depends on D19048. Fixes T11330.
Test Plan: Wrote rules to call webhooks selectively, saw them fire appropriately with correct trigger attribution.
Maniphest Tasks: T11330
Differential Revision: https://secure.phabricator.com/D19049
Summary: Depends on D19047. Ref T11330. Triggers every firehose hook on every edit; prepares for Herald triggers.
Test Plan: Configured a firehose hook, edited some objects, saw callbacks.
Maniphest Tasks: T11330
Differential Revision: https://secure.phabricator.com/D19048
Summary: Ref T13053. See PHI126. Add an explicit "Mute" action to kill mail and notifications for a particular object.
Test Plan: Muted and umuted an object while interacting with it. Saw mail route appropriately.
Maniphest Tasks: T13053
Differential Revision: https://secure.phabricator.com/D19033
Summary:
Ref T13053. Fixes T7804. Adds "Acting user" so you can have "always email me" stuff skip things you did or keep an eye on suspicious interns.
For the test console, the current user is the acting user.
For pushes, the pusher is the acting user.
Test Plan: Wrote acting user rules, triggered them via test console and via multiple actors on real objects.
Maniphest Tasks: T13053, T7804
Differential Revision: https://secure.phabricator.com/D19031
Summary:
Depends on D19019. Ref T13053. Fixes T12689. See PHI178.
Currently, if `@alice` resigns from a revision but `#alice-fan-club` is still a subscriber or reviewer, she'll continue to get mail. This is undesirable.
When users are associated with an object but have explicitly disengaged in an individal role (currently, only resign in audit/differential) mark them "unexpandable", so that they can no longer be included through implicit membership in a group (a project or package).
`@alice` can still get mail if she's a explicit recipient: as an author, owner, or if she adds herself back as a subscriber.
Test Plan:
- Added `@ducker` and `#users-named-ducker` as reviewers. Ducker got mail.
- Resigned as ducker, stopped getting future mail.
- Subscribed explicitly, got mail again.
- (Plus some `var_dump()` sanity checking in the internals.)
Reviewers: amckinley
Maniphest Tasks: T13053, T12689
Differential Revision: https://secure.phabricator.com/D19021
Summary:
Depends on D19018. Fixes T4776. Ref T13053. When you remove someone from an object's recipient list (for example, by removing them a reviewer, auditor, subscriber, owner or author) we currently do not send them mail about it because they're no longer connected to the object.
In many of these cases (Commandeer, Reassign) the actual action in the UI adds them back to the object somehow (as a reviewer or subscriber, respectively) so this doesn't actually matter. However, there's no recovery mechanism for reviewer or subscriber removal.
This is slightly bad from a policy/threat viewpoint since it means an attacker can remove all the recipients of an object "somewhat" silently. This isn't really silent, but it's less un-silent than it should be.
It's also just not very good from a human interaction perspective: if Alice removes Bob as a reviewer, possibly "against his will", he should be notified about that. In the good case, Alice wrote a nice goodbye note that he should get to read. In the bad case, he should get a chance to correct the mistake.
Also add a `removed(@user)` mail stamp so you can route these locally if you want.
Test Plan:
- Created and edited some different objects without catching anything broken.
- Removed subscribers from tasks, saw the final email include the removed recipients with a `removed()` stamp.
I'm not totally sure this doesn't have any surprising behavior or break any weird objects, but I think anything that crops up should be easy to fix.
Reviewers: amckinley
Subscribers: sophiebits
Maniphest Tasks: T13053, T4776
Differential Revision: https://secure.phabricator.com/D19019
Summary:
Depends on D19004. Ref T13053. Ref T12677. If the new `cluster.mailers` is configured, make use of it. Also use it in the Sengrid/Mailgun inbound stuff.
Also fix a bug where "Must Encrypt" mail to no recipients could fatal because no `$mail` was returned.
Test Plan: Processed some mail locally. The testing on this is still pretty flimsy, but I plan to solidify it in an upcoming change.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13053, T12677
Differential Revision: https://secure.phabricator.com/D19005
Summary:
Ref T13053. When you mention one object on another (or link two objects together with an action like "Edit Parent Revisions"), we write a transaction on each side to add the "alice added subtask X" and "alice added parent task Y" items to the timeline.
This behavior now causes problems in T13053 with the "Must Encrypt" flag because it prevents the flag from being applied to the corresponding "inverse edge" mail.
This was added in rP5050389f as a quick workaround for a fatal related to Editors not having enough data to apply Herald on mentions. However, that was in 2014, and since then:
- Herald got a significant rewrite to modularize all the rules and adapters.
- Editing got a significant upgrade in EditEngine and most edit workflows now operate through EditEngine.
- We generally do more editing on more pathways, everything is more modular, and we have standardized how data is loaded to a greater degree.
I suspect there's no longer a problem with just running Herald here, and can't reproduce one. If anything does crop up, it's probably easy (and desirable) to fix it.
This makes Herald fire a little more often: if someone writes a rule, mentioning or creating a relationship to old tasks will now make the rule act. Offhand, that seems fine. If it turns out to be weird, we can probably tailor Herald's behavior.
Test Plan:
I wasn't able to break anything:
- Mentioned a task on another task (original issue).
- Linked tasks with commits, mocks, revisions.
- Linked revisions with commits, tasks.
- Mentioned users, revisions, and commits.
- Verified that mail generated by creating links (e.g., Revision > Edit Tasks) now gets the "Must Encrypt" flag properly.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13053
Differential Revision: https://secure.phabricator.com/D18999
Summary:
Ref T13053. Adds revision stamps (status, reviewers, etc). Adds Herald rule stamps, like the existing X-Herald-Rules header.
Removes the "self" stamps, since you can just write a rule against `whatever(@epriestley)` equivalently. If there's routing logic around this, it can live in the routing layer. This avoids tons of self-actor, self-mention, self-reviewer, self-blocking-reviewer, self-resigned-reviewer, etc., stamps.
Use `natcasesort()` instead of `sort()` so that numeric values (like monograms) sort `9, 80, 700` instead of `700, 80, 9`.
Remove the commas from rendering since they don't really add anything.
Test Plan: Edited tasks and revisions, looked at mail stamps, saw stamps that looked pretty reasonable (with no more self stuff, no more commas, sorting numbers, and Herald stamps).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13053
Differential Revision: https://secure.phabricator.com/D18997
Summary: Ref T13053. Adds more mail tags with information available on the Editor object.
Test Plan: Banged around in Maniphest, viewed the resulting mail, all the stamps seemed to align with reality.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13053
Differential Revision: https://secure.phabricator.com/D18995
Summary:
Ref T10448. Currently, we use "mail tags" (in {nav Settings > Email Preferences}) to give users some ability to route mail. There are a number of major issues with this:
- It isn't modular and can't be extended by third-party applications.
- The UI is a giant mess of 5,000 individual settings.
- Settings don't map clearly to actual edits.
- A lot of stuff isn't covered by any setting.
This adds a new system, called "mail stamps", which is similar to "mail tags" but tries to fix all these problems.
I called these "stamps" because: stamps make sense with mail; we can't throw away the old system just yet and need to keep it around for a bit; we don't use this term for anything else; it avoids confusion with project tags.
(Conceptually, imagine these as ink stamps like "RETURN TO SENDER" or "FRAGILE", not actual postage stamps.)
The only real "trick" here is that later versions of this will need to enumerate possible stamps for an object and maybe all possible stamps for all objects in the system. This is why stamp generation is separated into a "template" phase and a "value" phase. In future changes, the "template" phase can be used on its own to generate documentation and typeaheads and let users build rules. This may need some more refinement before it really works since I haven't built any of that yet.
Also adds a preference for getting stamps in the header only (default) or header and body (better for Gmail, which can't route based on headers).
Test Plan:
Fiddled with preference, sent some mail and saw a "STAMPS" setting in the body and an "X-Phabricator-Stamps" header.
{F5411694}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T10448
Differential Revision: https://secure.phabricator.com/D18991
Summary: Depends on D18983. Ref T13053. Adds a new Herald action to activate the "must encrypt" flag and drop mail content.
Test Plan:
- Created a new Herald rule:
{F5407075}
- Created a "dog task" (woof woof, unsecure) and a "duck task" (quack quack, secure).
- Viewed mail for both in `bin/mail` and web UI, saw appropriate security/encryption behavior.
- Viewed "Must Encrypt" in "Headers" tab for the duck mail, saw why the mail was encrypted (link to Herald rule).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13053
Differential Revision: https://secure.phabricator.com/D18984
Summary: Depends on D18946. Ref T13051. Begins writing edge transactions as just a list of changed PHIDs.
Test Plan: Added, edited, and removed projects. Reviewed transaction record and database. Saw no user-facing changes but a far more compact database representation.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13051
Differential Revision: https://secure.phabricator.com/D18947
Summary:
See PHI307. Currently, when reviews undraft, we retroactively add in older activity to the mail ("alice created this revision...").
However, we don't add that activity to the mail tags, so the relevant tags (like "revision created") are dropped forever.
Instead, use the same set of transactions for both mail body and mail tag construction.
This should be obsoleted in the relatively near future by T10448, but it's a better/more correct behavior in general and we probably can't get rid of tags completely for a while.
Test Plan:
Applied patch, created a revision with builds, saw it auto-undraft after builds finished. Used `bin/mail list-outbound` and `bin/mail show-outbound` to see the mail. Verified that it included retroactive text ("created this revision") AND retroactive tags.
Note that the tag for "A new revision is created" is `DifferentialTransaction::MAILTAG_REVIEW_REQUEST` with literal value `differential-review-request`.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D18941
Summary:
Ref T13042. This adds a "silent" edit mechanism which suppresses feed stories, email, and notifications.
The other behaviors here are:
- The transactions are marked as "silent" so we can render a hint in the UI in the future to make it clear to users that they aren't missing email.
- If the editor uses Herald, mail rules are suppressed so they don't fire incorrectly (this mostly affects "the first time this rule matches, send me an email" rules: without this, they'd match "the first time" on the bulk edit, not send email, then never match again since they already matched).
- If the edit queues additional edits, those are applied silently too.
This doesn't (or, at least, shouldn't) actually change any behavior since you can't apply silent edits yet.
Test Plan:
Somewhat theoretical, since this isn't reachable yet. Should get meaningful testing in an upcoming change.
Did a bit of var_dump() / debug poking to attempt to verify that nothing too crazy is happening.
Viewed and edited objects, no changes in behavior.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13042
Differential Revision: https://secure.phabricator.com/D18882
Summary:
Ref T13042. This is a very, very old policy-violating option from yesteryear which supported build systems publishing updates by adding comments to revisions, without sending email about it.
Harbormaster has served this role for a long time and this is policy-violating in the general case (it allows attackers to act in secret).
Test Plan: Grepped for affected symbols.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13042
Differential Revision: https://secure.phabricator.com/D18881
Summary:
Depends on D18851. Ref T13035. After D18819, revision creation transactions may be split into two groups (if prototypes are enabled).
This split means we have two workers. The first worker doesn't publish feed stories or mail; the second one does.
Currently, both workers call `shouldPublishFeedStory()` before they queue, and then again after the daemons pull them out of the queue. However, the answer to this question can change.
Specifically, this happens:
- `arc` creates a revision.
- The first transaction group applies, creating the revision as a draft, and returns `false` from `shouldPublishFeedStory()`, and does not generate related PHIDs. It queues a daemon to send mail, expecting it not to publish a feed story.
- The second transaction group applies, promoting the revision to "needs review". Since the revision has promoted, `shouldPublishFeedStory()` now returns true. This editor generates related PHIDs and queues a daemon task, expecting it to send mail / publish feed.
- A few milliseconds pass.
- The first job gets pulled out of the daemon queue and worked on. It does not have any feed metadata because the object wasn't publishable when the job was queued -- but `shouldPublishFeedStory()` now returns true, so it tries to publish a story without any metadata available. Slightly bad stuff happens (see below).
- The second job gets pulled out of the daemon queue and worked on. This one has metadata and works fine.
The "slightly bad stuff" is that we publish an empty feed story with no references to any objects, then try to push it to hooks and other listeners. Since it doesn't have any references, it fails to load during the "push to external listeners" phase.
This is harmless but clutters the log and doesn't help anything.
Instead, cache the state of "are we publishing a feed story for this object?" when we queue the worker so it can't race.
Test Plan:
- Enabled prototypes.
- Disabled all Herald triggers for Harbormaster build plans.
- Ran `bin/phd debug task` in one window.
- Created a revision in a separate window.
- Before patch: saw "unable to load feed story" errors in the daemon log.
- After patch: no more "unable to load feed story" errors.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13035
Differential Revision: https://secure.phabricator.com/D18852
Summary:
Fixes T13027. Ref T2543. When revisions promote from "Draft" because builds finish or no builds are configured, the status currently switches from "Draft" to "Needs Review" without re-running Herald.
This means that some rules -- notably, "Send me an email" rules -- don't fire as soon as they should.
Instead of applying this promotion in a hacky way inline, queue it and apply it normally in a second edit, after the current group finishes.
Test Plan:
- Created a revision, reviewed Herald transcripts.
- Saw three Herald passes:
- First pass (revision creation) triggered builds and no email.
- Second pass (builds finished) did not trigger builds (no update) and did not trigger email (revision still a draft).
- Third pass (after promotion out of 'draft') did not trigger builds (no update) but did trigger email (revision no longer a draft).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13027, T2543
Differential Revision: https://secure.phabricator.com/D18819
Summary: Noticed a couple of typos in the docs, and then things got out of hand.
Test Plan:
- Stared at the words until my eyes watered and the letters began to swim on the screen.
- Consulted a dictionary.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: #blessed_reviewers, epriestley
Subscribers: epriestley, yelirekim, PHID-OPKG-gm6ozazyms6q6i22gyam
Differential Revision: https://secure.phabricator.com/D18693
Summary:
Ref T2543. This doesn't stand alone since mail still goes out normally, but gets this piece working: new revisions start as "Draft", then after updates if there are no builds they go into "Needs Review".
This should work in general because builds update revisions when they complete, to publish a "Harbormaster finished build yada yada" transaction. So either we'll un-draft immediately, or un-draft after the last build finishes.
I'll hold this until the mail and some other stuff (like UI hints) are in slightly better shape since I think it's probably too rough on its own.
Test Plan: Created revisions locally, saw them un-draft after builds.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T2543
Differential Revision: https://secure.phabricator.com/D18628
