Summary: Ref T13513. When a revision has inlines with edit suggestions, pressing "j" and "k" can incorrectly select the blocks inside the diffs inside the inlines.
Test Plan: Used "j" to cycle through changes in a revision with inline comments with edit suggestions, didn't get jumped into the suggestion diffs.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21283
Summary:
Ref PHI1749. Instead of opening files to the last unchanged line on either side of the change, open files to the "simple" line number of the selected block.
For inlines, this is the inline line number.
For blocks, this is the first new-file line number, or the first old-file line number if no new-file line number exists in the block.
This may not always be what the user is hoping for (we can't know what the state of their working copy is) but should produce more obvious behavior.
Test Plan:
- In Diffusion, used "Open in Editor" with and without line selections. Saw same behavior as before.
- Used "n" and "r" to leave an inline with the keyboard, saw same behavior as before.
- Used "\" and "Open in Editor" menu item to open a file with:
- Nothing selected or changeset selected (line: 1).
- An inline selected (line: inline line).
- A block selected (line: first line in block, per above).
Differential Revision: https://secure.phabricator.com/D21282
Summary: Ref T13276. Ref T13513. All readers and writers were removed more than a year ago; clean up the last remnants of this table.
Test Plan: Grepped for table references, found none.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13513, T13276
Differential Revision: https://secure.phabricator.com/D21281
Summary:
Ref T13513. Syntax highlighting is potentially expensive, and the changeset rendering pipeline can cache it. However, the cache is currently keyed ONLY by Differential changeset ID.
Destroy the existing cache and rebuild it with a more standard cache key so it can be used in a more ad-hoc way by inline suggestion snippets.
Test Plan: Used Darkconsole, saw cache hits and no more inline syntax highlighting for changesets with many inlines.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21280
Summary: Ref T13513. Inline comment context information is somewhat expensive to construct and can be cached. Add a readthrough cache on top of it.
Test Plan: Loaded a source code changeset with many inline comments, used Darkconsole to inspect query activity. Saw caches get populated. Updated cache key, saw caches regenerate. Browsed Diffusion, nothing looked broken.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21279
Summary: Ref T13513. For now, I'm not supporting inline edit suggestions in Diffusion, although it's likely not difficult to do so in the future. Clean up some of the code so that plain ol' inlines work the same way they did before.
Test Plan:
- Created, edited, reloaded, submitted inlines in Diffusion: familiar old behavior.
- Created, edited, reloaded, submitted inlines with suggestions in Differential: familiar new behavior.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21278
Summary: Ref T13513. When rendering an inline suggestion for display, use highlighting and diffing.
Test Plan: {F7495053}
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21277
Summary:
Ref T13513. This still has quite a few rough edges and some significant performance isssues, but appears to mostly work.
Allow reviewers to "Suggest Edit" on an inline comment and provide replacement text for the highlighted source.
Test Plan: Created, edited, reloaded, and submitted inline comments in various states with and without suggestion text.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21276
Summary: Ref T13513. Introduce a formal server-side content state object so the whole state can be saved and restored to the drafts table, read from the request, etc.
Test Plan: Created and edited inlines. Reloaded drafts with edits. Submitted normal and editing comments. Grepped for affected symbols.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21275
Summary:
Ref T13513. If your 10 most recently authored inlines have all been deleted, these queries can fail by overheating. This is silly and probably rarely happens outside of development.
For now, just let them overheat. This may create a false negative (incorrect "no draft" signal when the real condition is "drafts, but 10 most recent comments were deleted"). This could be sorted out later with a query mode like "executeAny()", perhaps.
Test Plan:
- Created and deleted 10 inlines.
- Submitted comments.
- Before: overheating fatal during draft flag generation.
- After: clean submission.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21274
Summary:
Ref T13513. Currently, all the inline code passes around strings to describe content. I plan to add background music, animation effects, etc., soon. To prepare for this change, make content a state object.
This does not change any user-visible behavior, it just prepares for content to become more complicated than a single string.
Test Plan: Created, edited, submitted, cancelled, etc., comments.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21273
Summary:
See <https://discourse.phabricator-community.org/t/how-to-override-localhost-localdomain-in-email-message-id/3876/>.
Currently, Phabricator generates a "Message-ID" only in a subset of cases (roughly: when the message is first-in-thread and we expect the thread may have more than one message).
In cases where it does not generate a message ID, it expects the SMTP server to generate one for it. Servers will generally do this, and some ONLY do this (that is, they ignore IDs from Phabricator and replace them). Thus, several pieces of configuration control whether Phabricator attempts to generate a "Message-ID" at all.
The PHPMailer code has fallback behavior which generates a "<random>@localhost.localdomain" message ID. This is never desirable and ignores Phabricator-level configuration that Message IDs should not be generated.
For now, remove this code: it is never the desired behavior and sometimes explicitly contradicts the intent of configuration.
Possibly, a better change may be to make Phabricator always generate a message ID in cases where it isn't forbidden from doing so by configuration. However, that's a more complicated change and it's not clear if/when it would produce better behavior, so start here for now.
Test Plan: Confirmed by affected user (see linked thread).
Differential Revision: https://secure.phabricator.com/D21272
Summary: See PHI1745. Under PHP7, errors raised as Throwable miss this "generic exception" logic and don't increment their failure count. Instead, treat any "Throwable" we don't recognize like any "Exception" we don't recognize.
Test Plan:
- Under PHP7, caused a worker task to raise a Throwable (e.g., call to undefined method, see D21270).
- Ran `bin/worker execute --id ...`.
- Before: worker failed, but did not increment failure count.
- After: worker fails and increments failure count as it would for other types of unknown error.
Differential Revision: https://secure.phabricator.com/D21271
Summary: See PHI1745. This callsite for "ChangesetParser" was not properly updated for recent changes.
Test Plan:
- Set `metamta.differential.inline-patches` to 100.
- Created a new revision with a small (<100 line) diff, with at least one reviewer.
- Ran `bin/phd debug` and observed outbound mail queue with `bin/mail list-outbound`.
- Before: fatal when trying to generate the inline changes for mail.
- After: clean mail generation.
Differential Revision: https://secure.phabricator.com/D21270
Summary: See PHI1743. If a build has no initiator PHID, the rendering pathway incorrectly tries to access a handle for it anyway.
Test Plan:
- Set a build to have no initiator PHID.
- Viewed the build plan for the build.
- Before: fatal when trying to access the `null` handle.
- After: clean build plan rendering.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Differential Revision: https://secure.phabricator.com/D21269
Summary: Ref T13513. The code which added "r" and "R" to the inline menu accidentally discarded the difference between the keystrokes.
Test Plan:
- Clicked an inline, pressed "r", got new empty inline (previously: inline with quote).
- Clicked an inline, pressed "R", got a new quoted inline.
- Repeated steps with the menu items, got the expected behaviors.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21268
Summary:
Fixes T13539. See that task for discussion and a reproduction case.
This algorithm currently counts "\ No newline at end of file" lines as though they were normal source lines. This can cause offset issues in the rare case that a diff contains two of these lines (for each side of the file) and has changes between them (because the last line of the file was modified between the diffs).
Instead, don't count "\" as a display line.
Test Plan:
- See T13539 and PHI1740.
- Before: got fatals on the "wild" diff and the synthetic simplified version.
- After: clean intradiff rendering in both cases.
Maniphest Tasks: T13539
Differential Revision: https://secure.phabricator.com/D21267
Summary:
Ref T13538. See PHI1739. Synthetic Git commits with no author and/or no commit message currently extract `null` and then fail to parse.
Ideally, we would carefully distinguish between `null` and empty string. In practice, that requires significant schema changes (these columns are non-nullable and have indexing requirements) and these cases are degenerate. These commits are challenging to build and can not normally be constructed with `git commit`.
At least for now, merge the `null` cases into the empty string cases so we can survive import.
Test Plan:
- Constructed a commit with no author and no commit message using the approach described in T13538; pushed and parsed it.
- Before: fatals during identity selection and storing the commit message (both roughly NULL inserts into non-null columns).
- After: clean import.
This produces a less-than-ideal UI in Diffusion, but it doesn't break anything:
{F7492094}
Maniphest Tasks: T13538
Differential Revision: https://secure.phabricator.com/D21266
Summary:
Fixes T13536. See that task for discussion.
Older versions of MySQL (roughly, prior to 8.0.19) emit "int(10)" types. Newer versions emit "int" types. Accept these as equivalent.
Test Plan: Ran `bin/storage upgrade --force` against MySQL 8.0.11 and 8.0.20. Got clean adjustment lists on both versions.
Maniphest Tasks: T13536
Differential Revision: https://secure.phabricator.com/D21265
Summary: On the "New User" web workflow, if you use an invalid email address, you get a failure with an empty message.
Test Plan:
- Before: Tried to create a new user with address "asdf". Got no specific guidance.
- After: Got specific guidance about email address formatting and length.
Differential Revision: https://secure.phabricator.com/D21264
Summary: Ref T13513. On `secure`, I caught an issue where inlines may be accessed directly before they're constructed. Instead, access them through the relevant accessor.
Test Plan: Will deploy.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21263
Summary:
Ref T13513. The on-hover-inline reticle has switched over to have cell-based behavior. Switch the on-hover-line-number reticle to use the same behavior.
Also, clean up the dirty/redraw loop slightly: we no longer need to dirty on resize, and we don't need to redraw if the range isn't actually dirty.
Test Plan: Highlighted lines and line ranges. Hovered over inlines.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21262
Summary:
Ref T13513. See PHI1734, which raises a concern about the performance of large revisions near the 100-change threshold.
Currently, `getInlines()` is called whenever the scroll position transitions between two changesets, and it performs a relatively complicated DOM scan to lift inlines out of the document.
This shows up as taking a small but nontrivial amount of time in Firefox profiles and should be safely memoizable.
Test Plan:
- Under Firefox profiling, scrolled through a large revision.
- Before change: `getInlines()` appeared as the highest-cost thing we're explicitly doing on profiles.
- After change: `getInlines()` was no longer meaningfully represented on profiles.
- Created inlines, edited inlines, etc. Didn't identify any broken behavior.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21261
Summary:
Ref T13529. Now that instances can be renamed, an instance may have multiple valid SSH usernames and the preferred SSH username may not be the intenal instance name.
`PhacilitySiteSource` should already always set `diffusion.ssh-username` correctly, to the current preferred SSH username (which may be "new-name" after a rename from "old-name"), so we should never be able to reach this code without an accurate `diffusion.ssh-username` value available.
The code to resolve names into instances also already works for both "ssh old-name@..." and "ssh new-name@...".
So I believe this code has no beneficial effects and only causes harm: it may force us to return "old-name" when falling through would correctly return "new-name".
Test Plan:
- Previously: renamed an instance, then SSH'd to it using both the old and new names. Both work.
- Previously: verified that `diffusion.ssh-username` is set correctly after a rename.
- Verified that Diffusion "Clone" UI now shows "new-name" after an instance rename.
- The real question here is: does this break something I'm not thinking of? And the change probably has to go to production to answer that.
Maniphest Tasks: T13529
Differential Revision: https://secure.phabricator.com/D21259
Summary:
Ref T13513. The way I'm highlighting lines won't work for Jupyter notebooks or other complex content blocks, and I don't see an obvious way to make it work that's reasonably robust.
However, we can just ignore the range behavior for complex content and treat the entire block as selected. This isn't quite as fancy as the source behavior, but pretty good.
Also, adjust unified diff behavior to work correctly with highlighting and range selection.
Test Plan:
- Used range selection in a Jupyter notebook, got reasonable behavior (range is treated as "entire block").
- Used range selection in a unified diff, got equivalent behavior to 2-up diffs.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21257
Summary:
Ref T13513. Give selected inlines a selection state and visual cues which are similar to the changeset selection state.
Also fix a couple of minor issues with select interactions and offset comments.
Test Plan: Selected inlines, saw obvious visual cues.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21256
Summary: Ref T13513. Currently, clicking inline or changeset headers eats the click events. This doesn't serve any clear purpose, and means these clicks do not clear text selections from the document, which is unusual.
Test Plan:
- Selected some text in a diff.
- Clicked a changeset header to select it.
- Before patch: text selection and context menu were retained.
- After patch: text selection was cleared and context menu was dismissed.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21255
Summary:
Ref T13513. In Safari, do this:
- view a 2-up diff with content on both sides;
- select more than one line on the right side; and
- use your mouse to click "New Inline Comment" in the context menu that pops up.
The mousedown event for the "New Inline Comment" click removes the "copy selection" behavior and creates a flash where both sides of the diff are selected.
This doesn't happen with (most) normal clicks because mouse down on a non-grabbable element clears the document selection.
To avoid this, don't reset the copy selection behavior if the user mouses down on an "<a />". This might not be robust, but seems simple and plausible as a fix.
Test Plan:
- See above.
- Before patch: flash of overbroad selection when clicking "New Inline Comment".
- After patch: no selection flash.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21254
Summary: Ref T13513. This fixes a bug where clicking a line number, then clicking "Cancel" causes the paths panel to briefly update with an extra inline comment counted on the file.
Test Plan:
- Clicked a line number.
- Typed some text.
- Clicked "Cancel".
- Before patch: paths panel flashes "1".
- After patch: paths panel stays stable.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21253
Summary:
Ref T13513. Currently, when creating an inline by selecting a line range, slightly careless handling leads to an inline with "0" offsets (by passing "undefined" to the server). This causes the block to highlight every line except the last one as fully bright, which is incorrect.
An inline with "0" offsets and an inline with no offsets are different. Be more careful about passing offsets around and rendering them.
Test Plan:
- Used the line numbers to add an inline to lines 4-8 of a change.
- Hovered the inline.
- Saw all four lines marked as "dull"-highlighted (previously: three bright lines, one dull line).
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21252
Summary:
Ref T13513. When a user selects a text range and uses "New Inline Comment" to create a comment directly from a range, store the offset information alongside the comment.
When hovering the comment, highlight the original range.
Test Plan: {F7480926, size=full}
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21250
Summary:
Ref T13513.
- Firefox represents multiple selected rows as a discontinuous range. Accommodate this.
- Unified diffs don't have a "copy" marker. Do something sort-of-reasonable for them.
Test Plan:
- Selected multiple lines of content in Firefox, got an option to add a comment.
- Selected content in unified mode, got an option to add a comment.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21249
Summary: Ref T13513. Support direct text selection for inlines. This is currently just an alternate way to get to the same place as using line numbers, but can preserve offset/range information in the future.
Test Plan:
- Selected some text, hit "c", clicked "New Inline Comment", got sensible comments on both sides of a diff in Safari, Chrome, and (with limitations) Firefox.
- Caveats: no unified support, doesn't work across lines in Firefox.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21248
Summary: Ref T11401. Ref T13513. This paves the way for more comment actions, particularly an edit-after-submit action.
Test Plan: Took all actions from menus, via mouse and via keyboard (where applicable).
Maniphest Tasks: T13513, T11401
Differential Revision: https://secure.phabricator.com/D21244
Summary: See <https://github.com/phacility/phabricator/pull/854>. In some situations, `line-break: anywhere` produces better behavior than `word-break: break-all`. It never appears to produce worse behavior.
Test Plan:
- Break behavior changes if a line contains "<span />" elements caused by syntax highlighting. This CSS adjustment only appears to apply to text with internal "<span />" elements.
- This specifically impacts certain internal breakpoints adjacent to punctuation, so the test case is highly specific. Generic test cases with latin word characters do not evidence any behavioral changes.
- This change appears to have no impact on Safari, which uses the better behavior in all cases.
- Before Patch: In Firefox and Chrome, this specific change breaks awkwardly. There is more room for text to fit on the broken line:
Firefox
{F7480567}
Chrome
{F7480568}
- After Patch: Firefox and Chrome break the line better. Here's Firefox:
{F7480569}
- Additional context:
Safari Behavior (Unchanged)
{F7480570}
Chrome with no highlighting (desirable behavior). Firefox does the same thing.
{F7480571}
Also tested other cases, which seem never-worse in any browser.
{F7480574}
Differential Revision: https://secure.phabricator.com/D21247
Summary:
Ref T13454. See <https://discourse.phabricator-community.org/t/newly-created-ssh-private-keys-with-passphrase-not-working-anymore/3883>.
After changes to distinguish between invalid and passphrase-protected keys, SSH private key management code incorrectly uses "-y ..." ("print public key") when it means "-p ..." ("modify input file, removing passphrase"). This results in the command having no effect, and Passphrase stores the raw input credential, not the stripped version.
We can't recover the keys because we don't store the passphrase, so no migration here is really possible. (We could add more code to detect this case, but it's presumably rare.)
Also, correct the behavior of the "Show Public Key" action: this is available for users who can see the credential and does not require edit permission.
Test Plan:
- Created a new credential with a passphrase, then showed the public key.
Maniphest Tasks: T13006, T13454
Differential Revision: https://secure.phabricator.com/D21245
Summary:
Ref T13513. Currently, viewing a Jupyter document, hidden context just gets a plain "* * *" facade with no way to expand it.
Support click-to-expand, like source changes.
Test Plan:
- Clicked to expand various Jupyter diffs.
- Clicked to expand normal source changes.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21243
Summary:
Ref T13513. If you leave an inline on line 20 of a Jupyter document, we currently render context around *raw* line 20, which is inevitably some unrelated piece of JSON.
Instead, drop this context. (Ideal behavior would be to render context around Jupyter block 20, but that's a whole lot of work.)
Test Plan:
- On Jupyter changes and normal source changes, made and submitted inline comments, then viewed text and HTML mail.
- Saw no context on Jupyter comments (instead of bad context), and unchanged behavior (useful context) on normal source changes.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21242
Summary:
Ref T13513. Currently, "View as Document Type..." lists every available engine.
This is hard to get completely right because we can't always rebuild the document ref accurately in the endpoint, but try harder to fake something reasonable.
Test Plan: Used "View as Document Type..." on Jupyter notebooks, was given "Jupyter" and "Source" as options.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21241
Summary:
Ref T13513. As part of inline metadata, save the document engine the change is being rendered with.
This will allow other parts of the UI to detect that an inline was created on a Jupyter notebook but is being rendered on raw source, or whatever else.
The immediate goal is to fix nonsensical inline snippet rendering in email on Jupyter notebooks.
Test Plan:
- Created inlines and replies on normal soure code, saw no document engine annotated in the database.
- Created inlines and replies on a Jupyter notebook rendered in Jupyter mode, saw "jupyter" annotations in the database.
- Swapped document engines between Jupyter and Source, etc.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21240
Summary:
Ref T13513. If an intradiff has at least one unchanged file ("hasSameEffectAs()") or more than 100 files ("Large Change"), we hit this block and don't upcast storage inlines to runtime inlines. I missed this in testing.
Add the conversion step.
Test Plan: Viewed an intradiff with at least one unchanged file and at least one inline comment, saw correct rendering instead of fatal.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21239
Summary: Ref T13513. Currently, if you're editing a comment, "delete" doesn't put the comment into the correct state. This action is normally only reachable from comment previews, since an editing inline has no "delete" button.
Test Plan:
- Started editing an inline, clicked "Delete", got a deletion.
- Created an inline, typed text,
- Deleted a normal comment via preview.
- Deleted a normal comment via the on-inline action.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21238
Summary:
Ref T13513. Currently, clicking "View" from the inline comment preview (below the "add comment" area at the bottom of the page) only works if the inline isn't being edited.
Update this behavior so it works on inlines in either "Viewing" or "Editing" states.
Test Plan:
- Clicked "View" on a normal inline, got jumped/selected.
- Clicked "View" on an editing inline, got jumped/selected.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21237
Summary:
Ref T13513. Currently:
- If you click the "Show Changeset" button, your state change doesn't actually get saved on the server.
- It's hard to select a changeset path name for copy/paste because the "highlight the header" code tends to eat the event.
Instead: persist the former event; make the actual path text not be part of the highlight hitbox.
Test Plan:
- Clicked "Show Changeset", reloaded, saw changeset visibility persisted.
- Selected changeset path text without issues.
- Clicked non-text header area to select/deselect changesets.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21236
Summary:
Ref T13513. Currently, if you:
- click a line to create an inline;
- type some text;
- wait a moment; and
- close the page.
...you don't get an "Unsubmitted Draft" marker in the revision list.
Lift all the draft behavior to "InlineController" and make saving a draft dirty the overall container draft state.
Test Plan:
- Took the steps described above, got a draft state marker.
- Created, edited, submitted, etc., inlines in Diffusion and Differential.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21235
Summary: Ref T13513. All queries now go through a reasonably minimal set of pathways and should have consistent behavior.
Test Plan:
- Loaded a revision with inlines.
- Created a new empty inline, reloaded page, saw it vanish.
- Created a new empty inline, typed draft text, did not save, reloaded page, saw draft present.
- Created a new empty inline, typed draft text. Submitted feedback, got prompt, answered "Y", saw draft text submit.
- Created a new empty inline, typed draft text, scrolled down to bottom of page, typed non-draft text, saw preview include draft text.
- Marked and submitted "Done".
- Used hide/show on inlines, verified state persisted.
- Did much of the same stuff in Diffusion, where it all works the same way (except: there's no prompt when submitting draft is-editing inlines).
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21234
Summary: Ref T13513. Replaces "DifferentialInlineCommentQuery" with the similar but more modern "DifferentialDiffInlineCommentQuery".
Test Plan: Viewed comments in timeline, changesets. Created, edited, and submitted comments. Hid and un-hid comments, reloading (saw state preserved).
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21233
Summary: Ref T13513. Continue removing usage sites for the obsolete "DifferentialInlineCommentQuery".
Test Plan: Viewed the inline list in Differential, saw sensible inlines.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21232
Summary: Ref T13513. Move querying to "DiffInlineCommentQuery" classes and lift them into the base Controller.
Test Plan: In Differential and Diffusion, created, edited, and submitted inline comments.
Maniphest Tasks: T13513
Differential Revision: https://secure.phabricator.com/D21231
