Summary: Fixes T13129. This at least makes the existing UI work again before we banish Phlux to the shadow realm.
Test Plan: Edited the visibility for a Phlux variable, didn't get an error. Nothing showed up in the edge tables when I made those changes, but at least it doesn't error out anymore.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin, PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13129
Differential Revision: https://secure.phabricator.com/D19387
Summary:
The name of networks should be unique.
Also adds support for exact-name queries for AlamanacNetworks.
Test Plan: Applied migration with existing duplicates, saw networks renamed, attempted to add duplicates, got a nice error message.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin, PHID-OPKG-gm6ozazyms6q6i22gyam
Differential Revision: https://secure.phabricator.com/D19379
Summary:
Depends on D19377. Ref T13125. Ref T13124. Ref T13105. Coverage reporting in Diffusion didn't initially survive the transition to Document Engine; restore it.
This adds some tentative/theoretical support for multiple columns of coverage, but no way to actually produce them in the UI. For now, the labels, codes, and colors are hard coded.
Test Plan:
Added coverage with `diffusion.updatecoverage`, saw coverage in the UI:
{F5525542}
Hovered over coverage, got labels and highlighting.
Double-checked labels for "N" (Not Executable) and "U" (Uncovered). See PHI577.
Faked some multi-column coverage, but you can't currently get this yourself today:
{F5525544}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13125, T13124, T13105
Differential Revision: https://secure.phabricator.com/D19378
Summary: Ref T13105. After moving Diffusion to DocumnentEngine, this no longer has callers. It will become part of the document behavior.
Test Plan: Grepped for calls to the `diffusion-browse-file` behavior, found none.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19377
Summary:
Depends on D19349. Ref T13105. This was the behavior in Diffusion before with a little hard-coded snippet.
Remove that snippet ("diffusion-jump-to") and add a more general-purpose snippet to SourceView.
This is a tiny bit hacky still (and probably doesn't work quite right with Quicksand) but gets things working again and works in all of Files, Paste, and Diffusion.
Test Plan: Followed links to particular lines in Paste, Files and Diffusion; got scrolled to the right place.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19350
Summary:
Depends on D19348. Ref T13105. When copying text from Paste or Diffusion, we'd like to copy only source, not line numbers.
We currently accomplish this with zero-width spaces plus a trigger that fires on "copy" in Paste and Diffusion. This is quite gross.
In the new-style Harbormaster logs, we use an approach that seems slightly better: CSS psuedoelements.
This isn't a complete solution (see also PHI504 / T5032) but puts us in a slightly better place.
Use it in Paste/Files/Diffusion too.
This gives us good behavior in all browsers in Files and Paste.
This gives us good behavior in Chrome and Firefox in Diffusion. Safari will copy (but not visually select) blame information in Diffusion. I think we can live with that for now.
Test Plan: Selected and copy/pasted stuff in Diffusion, Files, and Paste. Got good behavior everywhere except Safari + Diffusion.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19349
Summary:
Depends on D19347. Ref T13105. See PHI565. The "highlight lines" behavior is interacting poorly with the new blame element in Diffusion.
Make the behavior a little simpler and hopefully more robust.
Test Plan:
- Clicked commit/revision links in Diffusion, saw the links get followed instead of the lines highlighted.
- Highlighted lines in Diffusion, saw just the line/code highlight instead of the whole thing.
- Highlighted lines in Paste and new-style Harbormaster build logs, saw consistent behavior.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19348
Summary: Depends on D19346. Ref PHI568. I love Javascript.
Test Plan:
- Viewed a revision.
- Dragged file tree view really wide.
- Scrolled document to the right.
- Toggled file tree off and on by pressing "f" twice.
- Before patch: file tree grew wider and wider after it was toggled.
- After patch: file tree stayed the same size after it was toggled.
- Dragged to various widths and reloaded to make sure the "sticky across reloads" behavior still works.
- Scrolled right, dragged the tree a bit, then reloaded and didn't see it flip out.
Reviewers: amckinley
Reviewed By: amckinley
Differential Revision: https://secure.phabricator.com/D19347
Summary:
Ref T13105. Previously, the "source code" view in Paste rendered on a brown/orange-ish background. I've been using this element in more contexts (Files, Diffusion) and removed the colored background to make text (particularly syntax-highlighted text) easier to read and reduce visual noise with the new blame colors.
In Diffusion the view is in a box with a white background so removing the background left us with white, but in Paste it's just directly on the page so the background was bleeding through. Instead, set it to white explicitly.
Test Plan: Viewed source files in Files, Diffusion and Paste; saw text on a white background.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19346
Summary:
See PHI568. If you make the file tree UI very wide so that the page generates a horizontal scrollbar and then scroll the page, the page content can paint underneath the menu.
The menu already has a z-index to make it render above the content, but doesn't actually have a background. Give it a background.
The "transparent" rule was added in D16346 but I don't see any reason why we actually need it there, so I think this probably won't break anything.
Test Plan: {F5518822}
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13120
Differential Revision: https://secure.phabricator.com/D19344
Summary:
Depends on D19322. Ref T13120. Ref T12414.
Currently, `AlmanacDevice` has a bit of a beast of a `TYPE_INTERFACE` transaction that fully creates a complex Interface object. This isn't very flexible or consistent, and Interfaces are complex enough to reasonably have their own object behaviors (for example, they have their own PHIDs).
The complexity of this transaction makes modularizing `AlmanacDevice` transactions tricky. To simplify this, move Interface toward having its own set of normal transactions.
This change just adds some reasonable-looking transactions; it doesn't actually hook them up in the UI or make them reachable. I'll test that they actually work as I swap the UI over.
We may also have some code using the `TYPE_INTERFACE` transaction in Phacility support stuff, so that may need to wait a week to actually phase out.
Test Plan: Ran `bin/storage upgrade` and `arc liberate`. This code isn't reachable yet.
Reviewers: amckinley
Reviewed By: amckinley
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13120, T12414
Differential Revision: https://secure.phabricator.com/D19323
Summary: Ref T13118. The first fix there fixed Safari, but made Chrome weird. Try this?
Test Plan: Viewed a code block with `name=...` in Safari, Firefox and Chrome and saw consistent display without weird wrappping.
Reviewers: amckinley
Reviewed By: amckinley
Maniphest Tasks: T13118
Differential Revision: https://secure.phabricator.com/D19319
Summary:
Depends on D19313. Ref T13105. Fixes T13015. We lost the coloration for ages in the switch to Document Engine.
Restore it, and use a wider range of colors to make the information more clear.
Test Plan: Viewed some blame, saw a nice explosion of bright colors. This is a cornerstone of good design.
Maniphest Tasks: T13105, T13015
Differential Revision: https://secure.phabricator.com/D19314
Summary:
Depends on D19312. Ref T13105. For readability, render only one link for each contiguous block of changes.
Also make the actual rendering logic a little more defensible.
Test Plan: Viewed some files with blame, saw one render per chunk instead of one per line.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19313
Summary: Depends on D19311. Ref T13105. Currently, blame only renders on the initial request. Instead, redraw blame after swapping views.
Test Plan: Swapped from "Source -> Hexdump -> Source" and "Hexdump -> Source". Saw blame on source in all cases.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19312
Summary: Ref T13105. The line linker behavior currently has trouble identifying the line number when blame is active. Improve this, albeit not the most cleanly.
Test Plan: Selected lines with blame on.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19310
Summary: Ref T13105. This needs refinement but blame sort of works again, now.
Test Plan: Viewed files in Diffusion and Files; saw blame in Diffusion when viewing in source mode.
Reviewers: mydeveloperday
Reviewed By: mydeveloperday
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19309
Summary: Ref T13105. Ref T13047. This makes symbol indexes work with DocumentEngine in Files, and restores support in Diffusion.
Test Plan: Command-clicked stuff, got taken to the symbol index with reasonable metadata in Diffusion, Differential and Files.
Reviewers: mydeveloperday
Reviewed By: mydeveloperday
Maniphest Tasks: T13105, T13047
Differential Revision: https://secure.phabricator.com/D19307
Summary:
Ref T13105. Fixes some issues with line linking and highlighting under DocumentEngine:
- Adding `$1-3` to the URI didn't work correctly with query parameters.
- Reading `$1-3` from the URI didn't work correctly because Diffusion parses them slightly abnormally.
Test Plan: Clicked/dragged lines to select them. Observed URI. Reloaded page, got the right selection.
Reviewers: mydeveloperday
Reviewed By: mydeveloperday
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19305
Summary: Fixes T13118. Ref T13120. This construction is a little odd; I'm not entirely sure why Safari is doing what it's doing, but this appears to fix it.
Test Plan: Viewed blocks like those in T13118 in Safari. Before the patch, weird last-letter wrapping. After the patch, sensible behavior.
Maniphest Tasks: T13118, T13120
Differential Revision: https://secure.phabricator.com/D19303
Summary: Ref T13110. See PHI230. Show revision sizes on a roughly logarithmic scale from 1-7 stars. See D16322 for theorycrafting on this element.
Test Plan: Looked at some revisions, saw plausible-looking size markers.
Maniphest Tasks: T13110
Differential Revision: https://secure.phabricator.com/D19294
Summary:
Depends on D19284. Ref T13110. It's now possible to get a revision into a "Abandoned + But, Never Promoted From Draft" state. Show this in the header and provide the draft hint above the comment area.
Also, remove `shouldBroadcast()`. The method `getShouldBroadcast()` now has the same meaning.
Finally, migrate existing drafts to `shouldBroadcast = false` and default `shouldBroadcast` to `true`. If we don't do this, every older revision becomes a non-broadcasting revision because this flag was not explicitly set on revision creation before, only on promotion out of draft.
Test Plan: Ran migration; abandoned draft revisions and ended up in a draft + abandoned state.
Subscribers: PHID-OPKG-gm6ozazyms6q6i22gyam
Maniphest Tasks: T13110
Differential Revision: https://secure.phabricator.com/D19285
Summary: Depends on D19273. Ref T13105. Adds "Change Text Encoding..." and "Highlight As..." options when rendering documents, and makes an effort to automatically detect and handle text encoding.
Test Plan:
- Uploaded a Shift-JIS file, saw it auto-detect as Shift-JIS.
- Converted files between encodings.
- Highlighted various things as "Rainbow", etc.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19274
Summary:
Ref T13114. See PHI522. Although it looks like results are already ordered correctly, the override rendering isn't accommodating disabled results gracefully.
Give closed results a distinctive look (grey + strikethru) so it's clear when you're autocompleting `@mention...` into a disabled user.
Test Plan: {F5497621}
Maniphest Tasks: T13114
Differential Revision: https://secure.phabricator.com/D19272
Summary:
Depends on D19258. Ref T13105.
- When the default renderer is an Ajax renderer, don't replace the URI. For example, when viewing a Jupyter notebook, the URI should remain `/F123`, not instantly change to `/view/123/jupyter/`.
- Fix an issue where non-ajax renderers could fail to display the dropdown menu properly.
Test Plan:
- Viewed a Jupyter notebook, stayed on the same URI.
- Changed rendering, got different URIs.
- Viewed a JSON file and toggled renderers via dropdown.
Reviewers: mydeveloperday
Reviewed By: mydeveloperday
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19259
Summary:
Ref T13105. This adds various small client-side improvements to document rendering.
- In the menu, show which renderer is in use.
- Make linking to lines work.
- Make URIs persist information about which rendering engine is in use.
- Improve the UI feedback for transitions between document types.
- Load slower documents asynchronously by default.
- Discard irrelevant requests if you spam the view menu.
Test Plan: Loaded files, linked to lines, swapped between modes, copy/pasted URLs.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19256
Summary: Ref T13105. Allow normal text files to be rendered as documents, and add a "source code" rendering engine.
Test Plan: Viewed some source code.
Reviewers: mydeveloperday
Reviewed By: mydeveloperday
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19254
Summary:
Depends on D19252. Ref T13105. This very roughly renders Jupyter notebooks.
It's probably better than showing the raw JSON, but not by much.
Test Plan:
- Viewed various notebooks with various cell types, including markdown, code, stdout, stderr, images, HTML, and Javascript.
- HTML and Javascript are not live-fired since they're wildly dangerous.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19253
Summary:
Depends on D19251. Ref T13105. This adds rendering engine support for PDFs.
It doesn't actually render them, it just renders a link which you can click to view them in a new window. This is much easier than actually rendering them inline and at least 95% as good most of the time (and probably more-than-100%-as-good some of the time).
This makes PDF a viewable MIME type by default and adds a narrow CSP exception for it. See also T13112.
Test Plan:
- Viewed PDFs in Files, got a link to view them in a new tab.
- Clicked the link in Safari, Chrome, and Firefox; got inline PDFs.
- Verified primary CSP is still `object-src 'none'` with `curl ...`.
- Interacted with the vanilla lightbox element to check that it still works.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19252
Summary:
Ref T13105. Although Markdown is trickier to deal with, we can handle Remarkup easily.
This may need some support for encoding options.
Test Plan: Viewed `.remarkup` files, got remarkup document presentation by default. Viewed other text files, got an option to render as remarkup.
Reviewers: avivey
Reviewed By: avivey
Subscribers: mydeveloperday, avivey
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19251
Summary:
Depends on D19249. Ref T13109. Add timing information to the `PushEvent`:
- `writeWait`: Time spent waiting for a write lock.
- `readWait`: Time spent waiting for a read lock.
- `hostWait`: Roughly, total time spent on the leaf node.
The primary goal here is to see if `readWait` is meaningful in the wild. If it is, that motivates smarter routing, and the value of smarter routing can be demonstrated by looking for a reduction in read wait times.
Test Plan: Pushed some stuff, saw reasonable timing values in the table. Saw timing information in "Export Data".
Maniphest Tasks: T13109
Differential Revision: https://secure.phabricator.com/D19250
Summary:
Depends on D19247. Ref T13109. When we receive an SSH request, generate a random unique ID for the request. Then thread it down through the process tree.
The immediate goal is to let the `ssh-exec` process coordinate with `commit-hook` process and log information about read and write lock wait times. Today, there's no way for `ssh-exec` to interact with the `PushEvent`, but this is the most helpful place to store this data for users.
Test Plan: Made pushes, saw the `PushEvent` table populate with a random request ID. Exported data and saw the ID preserved in the export.
Maniphest Tasks: T13109
Differential Revision: https://secure.phabricator.com/D19249
Summary:
Depends on D19245. Fixes T11145. Ref T13108. See PHI488. Disable workflow buttons when they're clicked to prevent accidental client-side double submission.
This might have some weird side effects but we should normally never need to re-use a workflow dialog form so it's not immediately obvious that this can break anything.
Test Plan:
- Added `sleep(1)` to the Mute controller and the Maniphest task controller.
- Added `phlog(...)` to the Mute controller.
- Opened the mute dialog, mashed the button a thousand times.
- Before: Saw a bunch of logs.
- After: Button immediately disables, saw only one log.
Maniphest Tasks: T13108, T11145
Differential Revision: https://secure.phabricator.com/D19246
Summary:
See PHI488. Ref T13108. Currently, there is a narrow window between when the response returns and when the browser actually follows the redirect where the form is live and you can click the button again.
This is relativey easy if Phabricator is running //too fast// since the button may be disabled only momentarily. This seems to be easier in Firefox/Chrome than Safari.
Test Plan:
- In Firefox and Chrome, spam-clicked a comment submit button.
- Before: could sometimes get a double-submit.
- After: couldn't get a double-submit.
- This could probably be reproduced more reliabily by adding a `sleep(1)` to whatever we're redirecting //to//.
- Submitted an empty comment, got a dialog plus a still-enabled form (so this doesn't break the non-redirect case).
Maniphest Tasks: T13108
Differential Revision: https://secure.phabricator.com/D19245
Summary:
Depends on D19238. Ref T13105. Give document engines some reasonable automatic support for degrading gracefully when someone tries to hexdump a 100MB file or similar.
Also, make "Video" sort above "Audio" for files which could be rendered either way.
Test Plan: Viewed audio, video, image, and other files. Adjusted limits and saw full, partial, and fallback/error rendering.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19239
Summary: Depends on D19237. Ref T13105. This adds a (very basic) "Hexdump" engine (mostly just to have a second option to switch to) and a selector for choosing view modes.
Test Plan: Viewed some files, switched between audio/video/image/hexdump.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19238
Summary:
Ref T13105. This change begins modularizing document rendering. I'm starting in Files since it's the use case with the smallest amount of complexity.
Currently, we hard-coding the inline rendering for images, audio, and video. Instead, use the modular engine pattern to make rendering flexible and extensible.
There aren't any options for switching modes yet and none of the renderers do anything fancy. This API is also probably very unstable.
Test Plan: Viewwed images, audio, video, and other files. Saw reasonable renderings, with "nothing can render this" for any other file type.
Maniphest Tasks: T13105
Differential Revision: https://secure.phabricator.com/D19237
Summary:
See <https://discourse.phabricator-community.org/t/desktop-only-notifications-mode-is-broken/1234>. Ref T13102. The "Desktop Only" mode for notifications currently shows both desktop and web notifications.
In fact, `JX.Notification` currently has no ability to render notifications as desktop-only. Make this work.
Note that many of the variables and parameters here, including `showAnyNotification`, `web_ready`, and `desktop_ready`, are named in an incorrect or misleading way. However, the new behavior appears to be correct.
Test Plan:
- Emitted test notifications in "No Notifications", "Web Only", "Web and Desktop", and "Desktop" modes.
- Saw appropriate notifications appear in the UI.
Maniphest Tasks: T13102
Differential Revision: https://secure.phabricator.com/D19233
Summary: Ref T13106. When profiling service queries, there's no convenient way to easily get a sense of why a query was issued. Add a mode to collect traces for each query to make this more clear. This is rough, but works well enough to be useful.
Test Plan: Clicked "Analyze Query Plans", got stack traces for each service call.
Maniphest Tasks: T13106
Differential Revision: https://secure.phabricator.com/D19221
Summary:
See PHI451. Ref T13102. DarkConsole uses an ancient inline "onclick" handler to expand the stack traces for errors.
The new Content-Security-Policy prevents this from functioning.
Replace this with a more modern behavior-driven action instead.
Test Plan:
- Clicked some errors in DarkConsole, saw stack traces appear.
- Grepped for `onclick` and `jsprintf()` to see if I could find any more of these, but came up empty.
Maniphest Tasks: T13102
Differential Revision: https://secure.phabricator.com/D19218
Summary:
Fixes T12994. We need `MYSQLI_ASYNC` to implement client-side query timeouts, and we need MySQLi + MySQL Native Driver to get `MYSQLI_ASYNC`.
Recommend users install MySQLi and MySQL Native Driver if they don't have them. These are generally the defaults and best practice anyway, but Ubuntu makes it easy to use the older stuff.
All the cases we're currently aware of stem from `apt-get install php5-mysql` (which explicitly selects the non-native driver) so issue particular guidance about `php5-mysqlnd`.
Test Plan:
- Faked both issues locally, reviewed the text.
- Will deploy to `secure`, which currently has the non-native driver.
Maniphest Tasks: T12994
Differential Revision: https://secure.phabricator.com/D19216
Summary:
Ref T13102. An install has a custom rule for bridging JIRA references via Doorkeeper and would like to be able to render them as `JIRA-123` instead of `JIRA JIRA-123 Full JIRA title`.
I think it's reasonable to imagine future support upstream for `JIRA-123`, `{JIRA-123}`, and so on, although we do not support these today. We can take a small step toward eventual support by letting the rendering pipeline understand different view modes.
This adds an optional `name` (the default text rendered before we do the OAuth sync) and an optional `view`, which can be `short` or `full`.
Test Plan:
I tested this primarily with Asana, since it's less of a pain to set up than JIRA. The logic should be similar, hopefully.
I changed `DoorkeeperAsanaRemarkupRule` to specify `name` and `view`, e.g `'view' => (mt_rand(0, 1) ? 'short' : 'full')`. Then I made a bunch of Asana references in a comment and saw them randomly go short or long.
Maniphest Tasks: T13102
Differential Revision: https://secure.phabricator.com/D19215
Summary:
See PHI431. Ref T13102. An install is interested in a custom "non-sticky" accept action, roughly.
On the one hand, this is a pretty hacky patch. However, I suspect it inches us closer to T731, and I'm generally comfortable with exploring the realms of "Accept Next Update", "Unblock Without Accepting", etc., as long as most of it doesn't end up enabled by default in the upstream.
Test Plan:
- Accepted and updated revisions normally, saw accepts respect global stickiness.
- Modified the "Accept" action to explicitly be unsticky, saw nonsticky accept behavior after update.
Maniphest Tasks: T13102
Differential Revision: https://secure.phabricator.com/D19211
Summary: Ref T13103. Make favicons customizable, and perform dynamic compositing to add marker to indicate things like "unread messages".
Test Plan: Viewed favicons in Safari, Firefox and Chrome. With unread messages, saw pink dot composited into icon.
Maniphest Tasks: T13103
Differential Revision: https://secure.phabricator.com/D19209
Summary:
See PHI439. This fills in additional information about Owners packages.
Also removes dead `primaryOwnerPHID`.
Test Plan: Called `owners.search` and reviewed the results. Grepped for `primaryOwnerPHID`.
Differential Revision: https://secure.phabricator.com/D19207
Summary:
See <https://discourse.phabricator-community.org/t/lightbox-not-working/1201/3>. The lightbox code is fragile and currently relies on simulating a click on the actual "<a />" tag surrounding other images in the document.
This breaks the prev/next links which ignore the event because it there's no "<img />".
Instead, don't simulate clicks and just call the code we want directly.
Test Plan: Added several images to a page, used lightbox prev/next buttons to cycle between them.
Differential Revision: https://secure.phabricator.com/D19197
Summary:
Depends on D19192. Ref T4190. Ref T13101. Instead of directly including the proxy endpoint with `<img src="..." />`, emit a placeholder and use AJAX to make the request. If the proxy fetch fails, replace the placeholder with an error message.
This isn't the most polished implementation imaginable, but it's much less mysterious about errors.
Test Plan: Used `{image ...}` for valid and invalid images, got images and useful error messages respectively.
Maniphest Tasks: T13101, T4190
Differential Revision: https://secure.phabricator.com/D19193
Summary: Depends on D19190. Fixes T12590. Ref T13099. Replaces the barely-usable, gigantic, poorly ordered "<select />" control with a tokenizer. Attempts to fix various minor issues.
Test Plan:
- Edited paths: include/exclude paths, from different repositories, different actual paths.
- Used "Add New Path" to add rows, got repository selector prepopulated with last value.
- Used "remove".
- Used validation typeahead, got reasonable behaviors?
The error behavior if you delete the repository for a path is a little sketchy still, but roughly okay.
Maniphest Tasks: T13099, T12590
Differential Revision: https://secure.phabricator.com/D19191
