Summary:
Ref T13250. See PHI1069. This is a small fix for `getRequestURI()` currently not working if the request includes "x[]=..." PHP-flavored array parameters, beacause they're parsed into arrays by `$_GET` and `setQueryParams(...)` no longer accepts nonscalars.
Instead, just parse the raw request URI.
Test Plan: Visited `/search/hovercard/?phids[]=X`, no more fatal. Dumped the resulting URI, saw it had the right value. Tried `?phids[]=x&x=1&x=1&x=1`, saw the parameters correctly preserved.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250
Differential Revision: https://secure.phabricator.com/D20147
Summary: See D20136. This method is sort of inherently bad because it is destructive for some inputs (`x=1&x=2`) and had "PHP-flavored" behavior for other inputs (`x[]=1&x[]=2`). Move to explicit `...AsMap` and `...AsPairList` methods.
Test Plan: Bit of an adventure, see inlines in a minute.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20141
Summary:
Depends on D20140. Ref T13250. Currently, the top-level exception handler doesn't dump stacks because we might not be in debug mode, and we might double-extra-super fatal if we call `PhabricatorEnv:...` to try to figure out if we're in debug mode or not.
We can get around this by setting a flag on the Sink once we're able to confirm that we're in debug mode. Then it's okay for the top-level error handler to show traces.
There's still some small possibility that showing a trace could make us double-super-fatal since we have to call a little more code, but AphrontStackTraceView is pretty conservative about what it does and 99% of the time this is a huge improvement.
Test Plan: {F6205122}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250
Differential Revision: https://secure.phabricator.com/D20142
Summary:
Depends on D20137. Ref T13250. Ref T12101. In versions of PHP beyond 7, various engine errors are gradually changing from internal fatals or internal errors to `Throwables`, a superclass of `Exception`.
This is generally a good change, but code written against PHP 5.x before `Throwable` was introduced may not catch these errors, even when the code is intended to be a top-level exception handler.
(The double-catch pattern here and elsewhere is because `Throwable` does not exist in older PHP, so `catch (Throwable $ex)` catches nothing. The `Exception $ex` clause catches everything in old PHP, the `Throwable $ex` clause catches everything in newer PHP.)
Generalize some `Exception` into `Throwable`.
Test Plan:
- Added a bogus function call to the rendering stack.
- Before change: got a blank page.
- After change: nice exception page.
{F6205012}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250, T12101
Differential Revision: https://secure.phabricator.com/D20138
Summary:
Ref T13250. When exceptions occur in display/rendering/writing, they currently go straight to the fallback handler. This is a minimal handler which doesn't show a stack trace or include any debugging details.
In some cases, we have to do this: some of these exceptions prevent us from building a normal page. For example, if the menu bar has a hard fatal in it, we aren't going to be able to build a nice exception page with a menu bar no matter how hard we try.
However, in many cases the error is mundane: something detected something invalid and raised an exception during rendering. In these cases there's no problem with the page chrome or the rendering pathway itself, just with rendering the page data.
When we get a rendering/response exception, try a second time to build a nice normal exception page. This will often work. If it doesn't work, fall back as before.
Test Plan:
- Forced the error from T13250 by applying D20136 but not D20134.
- Before:
{F6205001}
- After:
{F6205002}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250
Differential Revision: https://secure.phabricator.com/D20137
Summary:
Depends on D20138. Ref T13250. This improves exception behavior and gives us a standard page with a stack trace instead of a text fatal with no stack trace.
Truly a great day for PHP.
(Eventually we may want to replace all `(string)` with `phutil_string_cast()`, but let's let it have some time in the wild first?)
Test Plan: Triggered the error, got a more useful exception behavior.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250
Differential Revision: https://secure.phabricator.com/D20140
Summary: See D20126. I was trying to be a little too cute here with the names and ended up confusing myself, then just tested the method behavior. :/
Test Plan: Persudaded by arguments in D20126.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20135
Summary:
Ref T13250. Currently, datasources have a `setParameters(...)` method. This method accepts a dictionary and adds the key/value pairs to the raw HTTP request to the datasource endpoint.
Since D20049, this no longer works. Since D20116, it fatals explicitly.
In general, the datasource endpoint accepts other values (like `query`, `offset`, and `limit`), and even before these changes, using secret reserved keys in `setParameters(...)` would silently cause program misbehavior.
To deal with this, pass parameters as a JSON string named "parameters". This fixes the HTTP query issue (the more pressing issue affecting users today) and prevents the "shadowing reserved keys" issue (a theoretical issue which might affect users some day).
(I may revisit the `phutil_build_http_querystring()` behavior and possibly let it make this work again, but I think avoiding the duplicate key issue makes this change desirable even if the querystring behavior changes.)
Test Plan:
- Used "Land Revision", selected branches.
- Configured a custom Maniphest "users" field, used the search typeahead, selected users.
- Manually browsed to `/typeahead/class/PhabricatorPeopleDatasource/?query=hi¶meters=xyz` to see the JSON decode exception.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13250
Differential Revision: https://secure.phabricator.com/D20134
Summary:
Ref T13244. See PHI1059. When you lock a task, users who can edit the task can currently override the lock by using "Edit Task" if they confirm that they want to do this.
Mark these edits with an emblem, similar to the "MFA" and "Silent" emblems, so it's clear that they may have bent the rules.
Also, make the "MFA" and "Silent" emblems more easily visible.
Test Plan:
Edited a locked task, overrode the lock, got marked for it.
{F6195005}
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: aeiser
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20131
Summary: Ref T13244. See PHI1052. Our error handling for Stripe errors isn't great right now. We can give users a bit more information, and a less jarring UI.
Test Plan:
Before (this is in developer mode, production doesn't get a stack trace):
{F6197394}
After:
{F6197397}
- Tried all the invalid test codes listed here: https://stripe.com/docs/testing#cards
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20132
Summary:
Depends on D20129. Ref T13244. See PHI1058. When a revision has an "Accept" from a package, count the owners as "involved" in the change whether or not any actual human owners are actually accepting reviewers.
If a user owns "/" and uses "force accept" to cause "/src/javascript" to accept, or a user who legitimately owns "/src/javascript" accepts on behalf of the package but not on behalf of themselves (for whatever reason), it generally makes practical sense that these changes have owners involved in them (i.e., that's what a normal user would expect in both cases) and don't need to trigger audits under "no involvement" rules.
Test Plan: Used `bin/repository reparse --force --owners <commit>` to trigger audit logic. Saw a commit owned by `O1` with a revision counted as "involved" when `O1` had accepted the revision, even though no actual human owner had accepted it.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20130
Summary:
Depends on D20124. Ref T13244. See PHI1055. Add a few more builtin audit behaviors to make Owners more flexible.
(At the upper end of flexibility you can trigger audits in a very granular way with Herald, but you tend to need to write one rule per Owners package, and providing a middle ground here has worked reasonably well for "review" rules so far.)
Test Plan:
- Edited a package to select the various different audit rules.
- Used `bin/repository reparse --force --owners <commit>` to trigger package audits under varied conditions.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20126
Summary:
Ref T13244. See PHI1055. (Earlier, see D20091 and PHI1047.) Previously, we expanded the Owners package autoreview rules from "Yes/No" to several "Review (Blocking) If Non-Owner Author Not Subscribed via Package" kinds of rules. The sky didn't fall and this feature didn't turn into "Herald-in-Owners", so I'm comfortable doing something similar to the "Audit" rules.
PHI1055 is a request for a way to configure slightly different audit behavior, and expanding the options seems like a good approach to satisfy the use case.
Prepare to add more options by moving everything into a class that defines all the behavior of different states, and converting the "0/1" boolean column to a text column.
Test Plan:
- Created several packages, some with and some without auditing.
- Inspected database for: package state; and associated transactions.
- Ran the migrations.
- Inspected database to confirm that state and transactions migrated correctly.
- Reviewed transaction logs.
- Created and edited packages and audit state.
- Viewed the "Package List" element in Diffusion.
- Pulled package information with `owners.search`, got sensible results.
- Edited package audit status with `owners.edit`.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20124
Summary:
Ref T13244. See PHI1057. Currently, if you're a member of a lot of projects/packages, you can end up with a very large `commit.authorPHID IN (...)` clause in part of the "Active Audits" query, since your `alice` token in "Responsible Users: alice" expands into every package and project you can audit on behalf of.
It's impossible for a commit to be authored by anything but a user, and evidence in PHI1057 suggests this giant `IN (...)` list can prevent MySQL from making effective utilization of the `<authorPHID, auditStatus, ...>` key on the table.
Prefilter the list of PHIDs to only PHIDs which can possibly author a commit.
(We'll also eventually need to convert the `authorPHIDs` into `identityPHIDs` anyway, for T12164, and this moves us slightly toward that.)
Test Plan: Loaded "Active Audits" before and after change, saw a more streamlined and sensible `authorPHID IN (...)` clause afterwards.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20129
Summary:
Depends on D20118. Fixes T5351. We possibly raise some warnings about registration (approval queue, email domains), but they aren't relevant if no one can register.
Hide these warnings if no providers actually support registration.
Test Plan: Viewed the Auth provider list with registration providers and with no registration providers, saw more tailored guidance.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T5351
Differential Revision: https://secure.phabricator.com/D20119
Summary:
Depends on D20122. Fixes T8029. Adds an "Approve User" action to the "Manage" page.
Users are normally approved from the "Approval Queue", but if you click into a user's profile to check them out in more detail it kind of dead ends you right now. I've occasionally hit this myself, and think this workflow is generally reasonable enough to support upstream.
Test Plan: {F6193742}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T8029
Differential Revision: https://secure.phabricator.com/D20123
Summary: Depends on D20120. Fixes T8907. I thought this needed some Javascript nonsense but Safari, Firefox and Chrome all support an `autofocus` attribute.
Test Plan: Loaded login page with password auth enabled in Safari, Firefox, and Chrome; saw username field automatically gain focus.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T8907
Differential Revision: https://secure.phabricator.com/D20122
Summary: Depends on D20108. Ref T6703. Update this outdated callsite to a more modern appraoch.
Test Plan: Destroyed a user with `bin/remove destroy`.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T6703
Differential Revision: https://secure.phabricator.com/D20109
Summary:
Depends on D20107. Ref T6703. Legalpad currently inserts "email" records into the external account table, but they're never used for anything and nothing else references them.
They also aren't necessary for anything important to work, and the only effect they have is making the UI say "External Account" instead of "None" under the "Account" column. In particular, the signatures still record the actual email address.
Stop doing this, remove all the references, and destroy all the rows.
(Long ago, Maniphest may also have done this, but no longer does. Nuance/Gatekeeper use a more modern and more suitable "ExternalObject" thing that I initially started adapting here before realizing that Legalpad doesn't actually care about this data.)
Test Plan: Signed documents with an email address, saw signature reflected properly in UI. Grepped for other callsites.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T6703
Differential Revision: https://secure.phabricator.com/D20108
Summary:
Depends on D20115. See <https://discourse.phabricator-community.org/t/transaction-search-endpoint-does-not-work-on-differential-diffs/2369/>.
Currently, `getApplicationTransactionCommentObject()` throws by default. Subclasses must override it to `return null` to indicate that they don't support comments.
This is silly, and leads to a bunch of code that does a `try / catch` around it, and at least some code (here, `transaction.search`) which doesn't `try / catch` and gets the wrong behavior as a result.
Just make it `return null` by default, meaning "no support for comments". Then remove the `try / catch` stuff and all the `return null` implementations.
Test Plan:
- Grepped for `getApplicationTransactionCommentObject()`, fixed each callsite / definition.
- Called `transaction.search` on a diff with transactions (i.e., not a sourced-from-commit diff).
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: jbrownEP
Differential Revision: https://secure.phabricator.com/D20121
Summary: Depends on D20114. This is on the way out, so make that explicitly clear.
Test Plan: Read setup issue and configuration option.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20115
Summary:
See <https://hackerone.com/reports/492525> and <https://hackerone.com/reports/489531>. I previously awarded a bounty for <https://hackerone.com/reports/434116> so Slowvote is getting "researched" a lot.
- Prevent users from undoing their vote by submitting the form with nothing selected.
- Prevent users from racing between the `delete()` and `save()` to vote for multiple options in a plurality poll.
Test Plan:
- Clicked the vote button with nothing selected in plurality and approval polls, got an error now.
- Added a `sleep(5)` between `delete()` and `save()`. Submitted different plurality votes in different windows. Before: votes raced, invalid end state. After: votes waited on the lock, arrived in a valid end state.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20125
Summary: Ref D20122. This is something I wanted in a bunch of places. Looks like at some point the most-annoying one (autofocus for entering TOTOP codes) already got fixed at some point.
Test Plan: Loaded the form, got autofocus as expected.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Differential Revision: https://secure.phabricator.com/D20128
Summary:
Depends on D20106. Ref T6703. Since I plan to change the `ExternalAccount` table, these migrations (which rely on `save()`) will stop working.
They could be rewritten to use raw queries, but I suspect few or no installs are affected. At least for now, just make them safe: if they would affect data, fatal and tell the user to perform a more gradual upgrade.
Also remove an `ALTER IGNORE TABLE` (this syntax was removed at some point) and fix a `%Q` when adjusting certain types of primary keys.
Test Plan: Ran `bin/storage upgrade --no-quickstart --force --namespace test1234` to get a complete migration since the beginning of time.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T6703
Differential Revision: https://secure.phabricator.com/D20107
Summary:
Depends on D20105. Fixes T7732. T7732 describes a case where a user had their Google credentials swapped and had trouble regaining access to their account.
Since we now allow email login even if password auth is disabled, it's okay to let users unlink their final account, and it's even reasonable for users to unlink their final account if it is mis-linked.
Just give them a warning that what they're doing is a little sketchy, rather than preventing the workflow.
Test Plan: Unlinked my only login account, got a stern warning instead of a dead end.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T7732
Differential Revision: https://secure.phabricator.com/D20106
Summary:
Ref T6703. Replaces the small "link" icon with a more obvious "Link External Account" button.
Moves us toward operating against `$config` objects instead of against `$provider` objects, which is more modern and will some day allow us to resolve T6703.
Test Plan: Viewed page, saw a more obvious button. Linked an external account.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T6703
Differential Revision: https://secure.phabricator.com/D20105
Summary: This option no longer needs to be configured if you configure inbound mail (and that's the easiest setup approach in a lot of cases), so stop telling users they have to set it up.
Test Plan: Read documentation and configuration help.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20104
Summary:
Depends on D20100. Ref T7732. Ref T13244. This is a bit of an adventure.
Long ago, passwords were digested with usernames as part of the salt. This was a mistake: it meant that your password becomes invalid if your username is changed.
(I think very very long ago, some other hashing may also have used usernames -- perhaps session hashing or CSRF hashing?)
To work around this, the "username change" email included a one-time login link and some language about resetting your password.
This flaw was fixed when passwords were moved to shared infrastructure (they're now salted more cleanly on a per-digest basis), and since D18908 (about a year ago) we've transparently upgraded password digests on use.
Although it's still technically possible that a username change could invalidate your password, it requires:
- You set the password on a version of Phabricator earlier than ~2018 Week 5 (about a year ago).
- You haven't logged into a version of Phabricator newer than that using your password since then.
- Your username is changed.
This probably affects more than zero users, but I suspect not //many// more than zero. These users can always use "Forgot password?" to recover account access.
Since the value of this is almost certainly very near zero now and declining over time, just get rid of it. Also move the actual mail out of `PhabricatorUser`, ala the similar recent change to welcome mail in D19989.
Test Plan: Changed a user's username, reviewed resulting mail with `bin/mail show-outbound`.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244, T7732
Differential Revision: https://secure.phabricator.com/D20102
Summary:
Depends on D20099. Ref T13244. See PHI774. When password auth is enabled, we support a standard email-based account recovery mechanism with "Forgot password?".
When password auth is not enabled, we disable the self-serve version of this mechanism. You can still get email account login links via "Send Welcome Mail" or "bin/auth recover".
There's no real technical, product, or security reason not to let everyone do email login all the time. On the technical front, these links already work and are used in other contexts. On the product front, we just need to tweak a couple of strings.
On the security front, there's some argument that this mechanism provides more overall surface area for an attacker, but if we find that argument compelling we should probably provide a way to disable the self-serve pathway in all cases, rather than coupling it to which providers are enabled.
Also, inch toward having things iterate over configurations (saved database objects) instead of providers (abstract implementations) so we can some day live in a world where we support multiple configurations of the same provider type (T6703).
Test Plan:
- With password auth enabled, reset password.
- Without password auth enabled, did an email login recovery.
{F6184910}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20100
Summary: Depends on D20096. Reverts D14057. This was added for Phacility use cases in D14057 but never used. It is obsoleted by {nav Auth > Customize Messages} for non-Phacility use cases.
Test Plan: Grepped for removed symbol.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20099
Summary:
Depends on D20095. Ref T13244. Currently, auth providers have a list item view and a single gigantic edit screen complete with a timeline, piles of instructions, supplemental information, etc.
As a step toward making this stuff easier to use and more modern, give them a separate view UI with normal actions, similar to basically every other type of object. Move the timeline and "Disable/Enable" to the view page (from the edit page and the list page, respectively).
Test Plan: Created, edited, and viewed auth providers.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20096
Summary:
Depends on D20094. Ref T13244. Ref T6703. See PHI774. Currently, we use an older-style radio-button UI to choose an auth provider type (Google, Password, LDAP, etc).
Instead, use a more modern click-to-select UI.
Test Plan: {F6184343}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244, T6703
Differential Revision: https://secure.phabricator.com/D20095
Summary:
Ref T13244. See PHI774. If an install does not use password auth, the "one-time login" flow (via "Welcome" email or "bin/auth recover") is pretty rough. Current behavior:
- If an install uses passwords, the user is prompted to set a password.
- If an install does not use passwords, you're dumped to `/settings/external/` to link an external account. This is pretty sketchy and this UI does not make it clear what users are expected to do (link an account) or why (so they can log in).
Instead, improve this flow:
- Password reset flow is fine.
- (Future Change) If there are external linkable accounts (like Google) and the user doesn't have any linked, I want to give users a flow like a password reset flow that says "link to an external account".
- (This Change) If you're an administrator and there are no providers at all, go to "/auth/" so you can set something up.
- (This Change) If we don't hit on any other rules, just go home?
This may be tweaked a bit as we go, but basically I want to refine the "/settings/external/" case into a more useful flow which gives users more of a chance of surviving it.
Test Plan: Logged in with passwords enabled (got password reset), with nothing enabled as an admin (got sent to Auth), and with something other than passwords enabled (got sent home).
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20094
Summary:
Ref T13244. This story publishes to the feed (and I think that's reasonable and desirable) but doesn't render as nicely as it could.
Improve the rendering.
(See T9233 for some context on why we render stories like this one in this way.)
Test Plan: {F6184490}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20097
Summary: See PHI1050. Although Diviner hasn't received a ton of new development for a while, it's: not exaclty new; and pretty useful for what we need it for.
Test Plan: Reading.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: leoluk
Differential Revision: https://secure.phabricator.com/D20086
Summary:
Ref T13244. See PHI1047. A while ago, the "Review" field changed from "yes/no" to 20 flavors of "Non-Owner Blocking Under A Full Moon". The sky didn't fall, so we'll probably do this to "Audit" eventually too.
The "owners.search" API method anticipates this and returns "none" or "audit" to describe package audit statuses, so it can begin returning "audit-non-owner-reviewers" or whatever in the future.
However, the "owners.edit" API method doesn't work the same way, and takes strings, and the strings have to be numbers. This is goofy and confusing and generally bad.
Make "owners.edit" take the same strings that "owners.search" emits. For now, continue accepting the old values of "0" and "1".
Test Plan:
- Edited audit status of packages via API using "none", "audit", "0", "1" (worked), and invalid values like "quack" (helpful error).
- Edited audit status of packages via web UI.
- Used `owners.search` to retrieve package information.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20091
Summary:
Ref T13244. See D20080. Rather than randomly jittering service calls, we can give each host a "metronome" that ticks every 60 seconds to get load to spread out after one cycle.
For example, web001 ticks (and makes a service call) when the second hand points at 0:17, web002 at 0:43, web003 at 0:04, etc.
For now I'm just planning to seed the metronomes randomly based on hostname, but we could conceivably give each host an assigned offset some day if we want perfectly smooth service call rates.
Test Plan: Ran unit tests.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13244
Differential Revision: https://secure.phabricator.com/D20087
Summary: The textarea is, in fact, above the description!
Test Plan: Description text changed.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: #blessed_reviewers, epriestley
Subscribers: Korvin, epriestley
Differential Revision: https://secure.phabricator.com/D20092
Summary: This seems generally reasonable, but is also a narrow fix to "Phacility scripts try to move instances into 'up', but the daemons can't MFA".
Test Plan: Launched a new instance locally, no more "daemons can't MFA" error.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20081
Summary: Ref T13242. Currently, the transaction query loads handles by default (this is unusual). We don't need them, so turn them off.
Test Plan: No apparent behavioral change, will compare production profiles.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13242
Differential Revision: https://secure.phabricator.com/D20068
Summary:
See PHI1015. If you add new repository nodes to a cluster, we may not actually sync some repositories for up to 6 hours (if they've had no commits for 30 days).
Add an explicit check for out-of-sync repositories to trigger background sync.
Test Plan:
- Ran `bin/phd debug pullocal`.
- Fiddled with the `repository_workingcopy` version table to put the local node in and out of sync with the cluster.
- Saw appropriate responses in the daemon (sync; wait if the last sync trigger was too recent).
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20078
Summary:
Fixes T13192. See PHI1015. When you deactivate a repository, we currently stop serving it.
This creates a problem for intracluster sync, since new nodes can't sync it. If nothing else, this means that if you "ship of theseus" your cluster and turn nodes over one at a time, you will eventually lose the entire repository. Since that's clearly a bad outcome, support sync.
Test Plan:
Testing this requires a "real" cluster, so I mostly used `secure`.
I deactivated rGITTEST and ran this on `secure002`:
```
./bin/repository thaw --demote secure002.phacility.net --force GITTEST && ./bin/repository update GITTEST
```
Before the patch, this failed:
```
[2019-01-31 19:40:37] EXCEPTION: (CommandException) Command failed with error #128!
COMMAND
git fetch --prune -- 'ssh://172.30.0.64:22/diffusion/GITTEST/' '+refs/*:refs/*'
STDOUT
(empty)
STDERR
Warning: Permanently added '172.30.0.64' (RSA) to the list of known hosts.
phabricator-ssh-exec: This repository ("rGITTEST") is not available over SSH.
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
```
After applying (a similar patch to) this patch to `secure001`, the sync worked.
I'll repeat this test with the actual patch once this deploys to `secure`.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13192
Differential Revision: https://secure.phabricator.com/D20077
Summary:
On instances, the "SiteSource" (for site config) pretty much copy-pastes the "read POST data" block because it needs to make some decisions based on POST data when handling inbound mail webhooks.
Move the upstream read a little earlier so we can get rid of this. Now that this step is separated and must happen before the profiler, there's no reason not to do it earlier.
Test Plan: POSTed some data across pages without issue, will remove duplicate code in upcoming change.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20073
Summary: See PHI1048. We have similar transactions elsewhere already (particularly, on tasks) and these edges don't have any special properties (like "Closes Txx As Wontfix" edges do) so this doesn't create any sort of peril.
Test Plan: {F6170556}
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D20075
