Summary:
The goal is to make fulltext search back-ends more extensible, configurable and robust.
When this is finished it will be possible to have multiple search storage back-ends and
potentially multiple instances of each.
Individual instances can be configured with roles such as 'read', 'write' which control
which hosts will receive writes to the index and which hosts will respond to queries.
These two roles make it possible to have any combination of:
* read-only
* write-only
* read-write
* disabled
This 'roles' mechanism is extensible to add new roles should that be needed in the future.
In addition to supporting multiple elasticsearch and mysql search instances, this refactors
the connection health monitoring infrastructure from PhabricatorDatabaseHealthRecord and
utilizes the same system for monitoring the health of elasticsearch nodes. This will
allow Wikimedia's phabricator to be redundant across data centers (mysql already is,
elasticsearch should be as well).
The real-world use-case I have in mind here is writing to two indexes (two elasticsearch clusters
in different data centers) but reading from only one. Then toggling the 'read' property when
we want to migrate to the other data center (and when we migrate from elasticsearch 2.x to 5.x)
Hopefully this is useful in the upstream as well.
Remaining TODO:
* test cases
* documentation
Test Plan:
(WARNING) This will most likely require the elasticsearch index to be deleted and re-created due to schema changes.
Tested with elasticsearch versions 2.4 and 5.2 using the following config:
```lang=json
"cluster.search": [
{
"type": "elasticsearch",
"hosts": [
{
"host": "localhost",
"roles": { "read": true, "write": true }
}
],
"port": 9200,
"protocol": "http",
"path": "/phabricator",
"version": 5
},
{
"type": "mysql",
"roles": { "write": true }
}
]
Also deployed the same changes to Wikimedia's production Phabricator instance without any issues whatsoever.
```
Reviewers: epriestley, #blessed_reviewers
Reviewed By: epriestley, #blessed_reviewers
Subscribers: Korvin, epriestley
Tags: #elasticsearch, #clusters, #wikimedia
Differential Revision: https://secure.phabricator.com/D17384
Under some workloads, the taskmaster may hibernate and launch more rapidly
than it should. Require 15 seconds of inactivity before hibernating. Also
hibernate for longer.
Auditors: chad
The `min()` vs `max()` fix in D17560 meant that the Trigger daemon only
hibernates for 5 seconds, so we do a full GC sweep every 5 seconds. This ends
up eating a fair amount of CPU for no real benefit.
The GC cursors should move to persistent storage, but just bump this default
up in the meantime.
Auditors: chad
If we try to render an edge transaction which uses unknown edge constants,
it turns out we fatal. Degrade instead. This happened when viewing very old
badges.
Auditors: chad
Summary: Ships Badges. I can write up some basic docs too if needed.
Test Plan: /applications/
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T12270
Differential Revision: https://secure.phabricator.com/D17360
Summary: Ref T12270. These no longer have any callsites.
Test Plan: Used `grep` to search for each edge class constant, found no hits.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12270
Differential Revision: https://secure.phabricator.com/D17562
Summary: Ref T12270. Adds a pager, plus a few little cleanups from copy/paste and accumulated cruft.
Test Plan:
- Paginated a user with 180 badges.
- Viewed a user with 0 badges.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12270
Differential Revision: https://secure.phabricator.com/D17561
Summary: Ref T12298. Like PullLocal daemons, this allows the last daemon in the pool to hibernate if there's no work to be done, and awakens the pool when work arrives.
Test Plan:
- Ran `bin/phd debug task --trace`.
- Saw the pool hibernate and look for tasks.
- Commented on an object.
- Saw the pool wake up and process the queue.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12298
Differential Revision: https://secure.phabricator.com/D17559
Summary:
Ref T11050. The old rule was "you can only resign if you're a reviewer".
With the new behavior of "resign", the rule should be "you can resign if you're a reviewer, or you have authority over any reviewer". Make it so.
Also fixes T12446. I don't know how to reproduce that but I'm pretty sure this'll fix it?
Test Plan:
- Could not resign from a revision with no authority/reviewer.
- Resigned from a revision with myself as a reviewer.
- Resigned from a revision with a package I owned as a reviewer.
- Could not resign from a revision I had already resigned from.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12446, T11050
Differential Revision: https://secure.phabricator.com/D17558
Summary:
Ref T12298. Two minor daemon improvements:
- Make the "waiting" message reflect hibernation.
- Don't trigger a reload right after launching.
Test Plan:
- Read "waiting" message.
- Ran "bin/phd start", didn't see an immediate SIGHUP in the log.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12298
Differential Revision: https://secure.phabricator.com/D17550
Summary: Fixes T9363. This drops empty buckets from dashboard panel context. Still see full results in Audit.
Test Plan: Create an "Active Audits" panel, add to Dashboard. See no commits found. Check Audit, see all buckets.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T9363
Differential Revision: https://secure.phabricator.com/D17545
Summary: Ref T9363, If we're in a dashboard panel, only show buckets with data, or a fallback if nothing exists.
Test Plan: Test 'active revisions' panel in a dashboard and in Differential.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T9363
Differential Revision: https://secure.phabricator.com/D17544
Summary: Ref T12298. This allows the PullLocal daemon to hibernate like the Trigger daemon, but automatically wakes it back up when it needs to do something.
Test Plan:
- Ran `bin/phd debug pulllocal --trace`.
- Saw the daemon hibernate after doing a checkup on repositories.
- Saw periodic queries to look for new update messages.
- After clicking "Update Now" in the web UI to schedule an update, saw the daemon wake up immediately.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12298
Differential Revision: https://secure.phabricator.com/D17540
Summary:
Ref T12444. A few issues:
- `x % (y - z)` doesn't generate values in the full range: the largest value is never generated. Instead, use `x % (1 + y - z)`.
- `digestToRange(1, count)` never generates 0. After fixing the first bug, it could generate `count`. The range of the arrays is `0..(count-1)`, inclusive. Generate the correct range instead.
- `unpack('L', ...)` can unpack a negative number on a 32-bit system. Use `& 0x7FFFFFFF` to mask off the sign bit so the result is always a positive integer.
- FileFinder might return arbitrary keys, but we rely on sequential keys (0, 1, 2, ...)
Test Plan:
- Used `bin/people profileimage ... --force` to regenerate images.
- Added some debugging to verify that the math seemed to be working.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12444
Differential Revision: https://secure.phabricator.com/D17543
Summary:
Fixes T12369. When you create objects they may technically be locked: either because the default state is legitimately locked, or because the default policies prevent you from viewing so we sort of technically end in a locked state.
Regardless, don't prompt during creation, since this prompt isn't useful even if the lock detection is completely legitimate.
Test Plan:
- In {nav Applications > Maniphest > Configure}, set "Default View Policy" to "No One".
- Tried to create a task.
- Before patch: prompted to override lock.
- After patch: no override prompt.
Reviewers: chad
Reviewed By: chad
Subscribers: d.maznekov
Maniphest Tasks: T12369
Differential Revision: https://secure.phabricator.com/D17541
Summary:
Ref T12271. Don't do anything with this yet, but store who accepted/rejected/whatever on behalf of reviewers.
In the future, we could use this to render stuff like "Blessed Committers (accepted by epriestley)" or whatever. I don't know that this is necessarily super useful, but it's easy to track, seems likely to be useful, and would be a gigantic pain to backfill later if we decide we want it.
Test Plan: Accepted/rejected a revision, saw reviewers update appropriately.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12271
Differential Revision: https://secure.phabricator.com/D17537
Summary:
Ref T12271. Currenty, when you "Accept" a revision, you always accept it for all reviewers you have authority over.
There are some situations where communication can be more clear if users can accept as only themselves, or for only some packages, etc. T12271 discusses some of these use cases in more depth.
Instead of making "Accept" a blanket action, default it to doing what it does now but let the user uncheck reviewers.
In cases where project/package reviewers aren't in use, this doesn't change anything.
For now, "reject" still acts the old way (reject everything). We could make that use checkboxes too, but I'm not sure there's as much of a use case for it, and I generally want users who are blocking stuff to have more direct accountability in a product sense.
Test Plan:
- Accepted normally.
- Accepted a subset.
- Tried to accept none.
- Tried to accept bogus reviewers.
- Accepted with myself not a reviewer
- Accepted with only one reviewer (just got normal "this will be accepted" text).
{F4251255}
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12271
Differential Revision: https://secure.phabricator.com/D17533
Summary: Hit this while `arc diff`'ing something which is triggering 2+ rules which add reviewers, I think.
Test Plan: Dug this out of a production stack trace; will push and `arc diff` again.
Reviewers: chad
Reviewed By: chad
Differential Revision: https://secure.phabricator.com/D17534
Summary: Fixes T12439. This pathway was just missing a `setContinueOnMissingFields(...)` to skip enforcement of required fields.
Test Plan:
- Added a required custom field.
- Mentioned any task without a field value in a comment.
- Edited that comment.
- Saved changes.
- Before fix: fatal in log.
- After fix: clean edit.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12439
Differential Revision: https://secure.phabricator.com/D17536
Summary: Fixes T11865. Part of a 'clean up remarkup' pass, removing Aleo helps simplify coding, is lighter on the wire, and gives a more consistent, clean look.
Test Plan: run celerity, grep for 'aleo' and 'Aleo', test Phriction, tasks
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T11865
Differential Revision: https://secure.phabricator.com/D17535
Summary:
Ref T10967. I'm not 100% sure we need this, but the old edge table had it and I recall an issue long ago where not having this key left us with a bad query plan.
Our data doesn't really provide a way to test this key (we have many revisions and few reviewers, so the query planner always uses revision keys), and building a convincing test case would take a while (lipsum needs some improvements to add reviewers). But in the worst case this key is mostly useless and wastes a few MB of disk space, which isn't a big deal.
So I can't conclusively prove that this key does anything to the dashboard query, but the migration removed it and I'm more comfortable keeping it so I'm not worried about breaking stuff.
At the very least, MySQL does select this key in the query plan when I do a "Reviewers:" query explicitly so it isn't //useless//.
Test Plan: Ran `bin/storage upgrade`, ran dashboard query, the query plan didn't get any worse.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17532
Summary:
Fixes T11050. Today, when a user resigns, we just delete the record of them ever being a reviewer.
However, this means you have no way to say "I don't care about this and don't want to see it on my dashboard" if you are a member of any project or package reviewers.
Instead, store "resigned" as a distinct state from "not a reviewer", and treat it a little differently in the UI:
- On the bucketing screen, discard revisions any responsible user has resigned from.
- On the main `/Dxxx` page, show these users as resigned explicitly (we could just hide them, too, but I think this is good to start with).
- In the query, don't treat a "resigned" state as a real "reviewer" (this change happened earlier, in D17517).
- When resigning, write a "resigned" state instead of deleting the row.
- When editing a list of reviewers, I'm still treating this reviewer as a reviewer and not special casing it. I think that's sufficiently clear but we could tailor this behavior later.
Test Plan:
- Resigned from a revision.
- Saw "Resigned" in reviewers list.
- Saw revision disappear from my dashboard.
- Edited revision, saw user still appear as an editable reviewer. Saved revision, saw no weird side effects.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11050
Differential Revision: https://secure.phabricator.com/D17531
Summary: Fixes T12434. I accidentally copy/pasted this too much in D17442.
Test Plan: Viewed a form edit page, no longer saw two copies of this action.
Reviewers: chad, cspeckmim
Reviewed By: chad, cspeckmim
Maniphest Tasks: T12434
Differential Revision: https://secure.phabricator.com/D17530
Summary: Ref T10390. Catch if the user doesn't have any dashboards they can edit and give them a helpful message instead.
Test Plan: Clean install, no dashboards, Click "Add to Dashboard" on ApplicationSearch results, see no dashboards message
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T10390
Differential Revision: https://secure.phabricator.com/D17528
Summary: Ref T10390. Dashboard usability is high enough that I think we should pin it by default for users to create custom home pages.
Test Plan: Review order of applications in sandbox.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T10390
Differential Revision: https://secure.phabricator.com/D17527
Summary: Ref T10390. Fixes the missing "fa-dashboard" icon and adds a few more for an even 25.
Test Plan: Create new dashboard, see dashboard icon, select new dashboard icon.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T10390
Differential Revision: https://secure.phabricator.com/D17526
Summary: Ref T10390. I find myself wanting to find dashboards I can edit, even if I am not the author. I think this is useful for larger installs with multiple admins. Also make disabled Dashboards more grey in UI results.
Test Plan: Log in a test user, create a dashboard with I cannot edit. Log into my account, search for editable dashboards and only see mine. Set dashboard to all users, search under test account and see editable dashboards.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T10390
Differential Revision: https://secure.phabricator.com/D17524
Summary: Ref T5307. Just makes the dialog a little easier to use. Picks a name if we already have one.
Test Plan: Test a builtin, custom saved, and a new advanced search (no name).
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T5307
Differential Revision: https://secure.phabricator.com/D17523
Summary:
Ref T10967. Improves some method names:
- `Revision->getReviewerStatus()` -> `Revision->getReviewers()`
- `Revision->attachReviewerStatus()` -> `Revision->attachReviewers()`
- `Reviewer->getStatus()` -> `Reviewer->getReviewerStatus()` (this is mostly to make this more greppable)
Test Plan:
- bunch o' `grep`
- Browsed around.
- If I missed anything, it should fatal in an obvious way. We have a lot of other `getStatus()` calls and it's hard to be sure I got them all.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17522
Summary: Ref T10967. The old name was because we had a `getReviewers()` tied to `needRelationships()`, rename this method to use a simpler and more clear name.
Test Plan: `grep`, browsed around.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17519
Summary:
Ref T10967. There have been two different ways to load reviewers for a while: `needReviewerStatus()` and `needRelationships()`.
The `needRelationships()` stuff was a false start along time ago that didn't really go anywhere. I believe the idea was that we might want to load several different types of edges (subscribers, reviewers, etc) on lots of different types of objects. However, all that stuff pretty much ended up modularizing so that main `Query` classes did not need to know about it, so `needRelationships()` never got generalized or went anywhere.
A handful of things still use it, but get rid of them: they should either `needReviewerStatus()` to get reviewer info, or the ~3 callsites that care about subscribers can just load them directly.
Test Plan:
- Grepped for removed methods (`needRelationships()`, `getReviewers()`, `getCCPHIDs()`, etc).
- Browsed Diffusion, Differential.
- Called `differential.query`.
It's possible I missed some stuff, but it should mostly show up as super obvious fatals ("call needReviewerStatus() before getReviewerStatus()!").
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17518
Summary:
Ref T10967.
When we query for revisions with particular reviewers, use the new table to drive the query.
When we load revisions for use in the application, also use the new table to drive the query.
This doesn't convert everything: there's some old `loadRelationships()` stuff still using the old table. But this moves the major stuff over.
(This also changes the icon for "commented" from a question mark to a speech bubble.)
Test Plan:
- Viewed revision lists and detail views on old and new code, saw identical outcomes.
- Updated revisions, accepted/rejected/commented on revisions.
- Hit the "Accepted Older" and "Commented Older" states by taking an action and then updating.
- Grepped for removed methods (like `getEdgeData()` and `getDiffID()`).
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17517
Summary:
Ref T10967. We still have double writes, so all reviewers are being written to both old and new storage. This migrates all the data in the old storage to the new storage, so both storage tables should have a complete set of data and be getting identical updates as we move forward.
After this, I can move readers over one at a time and eventually get rid of the old writes and old storage.
This loads all of the edge data into memory in a big chunk. I reached out to one install to get some more information about their data size. Ours is quite manageable and I think even large installs will probably fit into memory, but we can do this in chunks if not.
However, because the Edge table doesn't have an `id` column, we can't use either the `RawMigrationIterator` or the `MigrationIterator`, and would need to write a new `EdgeMigrationIterator`. This isn't tons of work but might not be necessary.
Test Plan: Ran the migration locally, spot-checked the results in the database for sanity and correctness.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17515
Summary:
Ref T10967. We have a "commented" state to help reviewers get a better sense of who is part of a discussion, and a "last action" state to help distinguish between "accept" and "accepted an older version", for the purposes of sticky accepts and as a UI hint.
Currently, these are first-class states, partly beacuse we were somewhat limited in what we could do with edges. However, a more flexible way to represent them is as flags separate from the primary state flag.
In the new storage, write them as separate state information: `lastActionDiffPHID` stores the Diff PHID of the last review action (accept, reject, etc). `lastCommentDiffPHID` stores the Diff PHID of the last comment (top-level or inline).
Test Plan: Applied storage changes, commented and acted on a revision. Saw appropriate state reflected in the database.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17514
Summary:
Ref T10967. `differential.createcomment` is a frozen API method which has been obsoleted by `differential.revision.edit`.
It is the only remaining way to apply an "accept", "reject", or "resign" action using the old "ACTION" code.
Instead of using the old code, sneakly apply a new type of transaction in these cases instead.
Then, remove all the remaining old code for this stuff on the write pathways.
Test Plan:
- Used "differential.createcomment" to accept, reject, and resign from a revision.
- Grepped for all removed ACTION_X constants, found them only in rendering code.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17513
Summary: Ref T10967. See that task for some discussion. This lets us do double writes on this pathway.
Test Plan: Set an Owners package to auto-review. Created revisions which triggered it: one with no reviewers (autoreview added); one with the package as a blocking reviewer explicitly (no automatic stuff happened, as expected).
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17512
Summary:
Ref T10967. This converts the reviewer update action in Herald from an older edge write to a newer ModularTransactions write.
The major value from this is that we get a double-write to the new reviewers table.
Test Plan:
- Wrote a Herald rule to add a reviewer and a blocking reviewer.
- Saw them added properly to a revision with: no reviewers; both as blocking; A as blocking, B as nonblocking; A as nonblocking, B as blocking.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10967
Differential Revision: https://secure.phabricator.com/D17511
Summary: Ref T5307. Actually check the built in query with query, not engine.
Test Plan: Try a builtin query, and a custom query when making a dashboard panel.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T5307
Differential Revision: https://secure.phabricator.com/D17521
Summary: Ref T5307. Adds a better query check query, sets required for the name, adds the correct URI for cancelling.
Test Plan: Test a form without a name, fake a query string, test cancel button.
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: Korvin
Maniphest Tasks: T5307
Differential Revision: https://secure.phabricator.com/D17520
Summary: Ref T5307. This adds an additional action to Use Results for creating a panel from the query.
Test Plan:
Navigate to Maniphest, select dropdown for Use Results. Try any of the following:
- Try to set a panel without a name (fail)
- Muck up query or engine (fail)
- Set a fake Dashboard ID (fail)
Give panel a name and select a dashboard I have edit permissions to, get taken to dashboard.
Reviewers: epriestley
Subscribers: Korvin
Maniphest Tasks: T5307
Differential Revision: https://secure.phabricator.com/D17516
Summary: Fixes T12418. This is a fairly advanced feature and I think users can reasonably consult the documentation for their own editors to figure out how to do this.
Test Plan: Saw no more text.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12418
Differential Revision: https://secure.phabricator.com/D17510
Summary: Fixes T12416. See that task for discussion. Slightly older versions of `git` do not appear to support use of `--` to separate flags and arguments.
Test Plan:
- Ran `bin/repository update PHABX`.
- In T12416, had a user with Git 2.1.4 confirm that `git ls-remote X` worked while `git ls-remote -- X` failed.
- Read `git help ls-remote` to look for any kind of suspicious `--destroy-the-world` flags, didn't see any that made me uneasy.
Reviewers: chad, avivey
Reviewed By: avivey
Maniphest Tasks: T12416
Differential Revision: https://secure.phabricator.com/D17508
Summary:
Ref T5378. This repackages an existing check to see if a URI is a URI for the current install into a more reasonable form.
In an upcoming change, I'll use this new check to test whether `http://example.whatever.com/T123` is a link to a task on the current install or not.
Test Plan: This stuff has good test coverage already; added some more.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T5378
Differential Revision: https://secure.phabricator.com/D17502
Summary:
Ref T12392. The logic currently goes like this:
- Try a fetch.
- If that fails, try repairing the origin URI.
- Then try again.
This is pretty complicated, and we can use this simpler logic instead:
- Set the origin URI to the right value.
- Try a fetch.
Setting the origin URI is very fast. This can normally only get us in any trouble in very obscure situations which haven't occurred for many years:
- Pretty much all of this is already covered by `verifyGitOrigin()`, which we run earlier.
- Origins could be configured to have multiple URIs for some reason, but shouldn't be.
- Years ago, you could configure Phabricator to point at a local repository it didn't own and that could conceivably have a different "origin" that you might not want us to delete. If you did this, the daemons have been spewing errors for 3-4 years without you fixing it. The cost of fixing the remote URI is very small even if anyone is affected by this (just set it back to the old value) and there's zero reason to do this and the scenario is ridiculous.
Test Plan: Ran `bin/repository update PHABX --trace --verbose`, saw fetches go through cleanly after URI adjustment.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12392
Differential Revision: https://secure.phabricator.com/D17498
Summary:
Ref T12296. Ref T12392. Currently, when we're observing a remote repository, we periodically run `git fetch ...`.
Instead, periodically run `git ls-remote` (to list refs in the remote) and `git for-each-ref` (to list local refs) and only continue if the two lists are different.
The motivations for this are:
- In T12296, it appears that doing this is //faster// than doing a no-op `git fetch`. This effect seems to reproduce locally in a clean environment (900ms for `ls-remote` + 100ms for `for-each-ref` vs about 1.4s for `fetch`). I don't have any explanation for why this is, but there it is. This isn't a huge change, although the time we're saving does appear to mostly be local CPU time, which is good for us.
- Because we control all writes, we could cache `git for-each-ref` in the future and do fewer disk operations. This doesn't necessarily seem too valuable, though.
- This allows us to tell if a fetch will do anything or not, and make better decisions around clustering (in particular, simplify how observed repository versioning works). With `git fetch`, we can't easily distinguish between "fetch, but nothing changed" and "legitimate fetch".
If a repository updates very regularly we end up doing slightly more work this way (that is, if `ls-remote` always comes back with changes, we do a little extra work), but this is normally very rare.
This might not get non-bare repositories quite right in some cases (i.e., incorrectly detect them as changed when they are unchanged) but we haven't created non-bare repositories for many years.
Test Plan: Ran `bin/repository update --trace --verbose PHABX`, saw sensible construction of local and remote maps and accurate detection of whether a fetch would do anything or not.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12392, T12296
Differential Revision: https://secure.phabricator.com/D17497
