3
0
mirror of https://github.com/ergochat/ergo.git synced 2024-11-14 07:59:31 +01:00
ergo/DEVELOPING.md
2018-02-11 05:30:40 -05:00

6.4 KiB
Raw Blame History

Developing Oragono

Most development happens on the develop branch, which is occasionally rebased + merged into master when its not incredibly broken. When this happens, the develop branch is usually pruned until I feel like making unsafe changes again.

I may also name the branch develop+feature if Im developing multiple, or particularly unstable, features.

The intent is to keep master relatively stable.

Releasing a new version

  1. Ensure dependencies are up-to-date.
  2. Run irctest over it to make sure nothings severely broken.
  3. Remove -unreleased from the version number in irc/constants.go.
  4. Update the changelog with new changes.
  5. Remove unused sections from the changelog, change the date/version number and write release notes.
  6. Commit the new changelog and constants change.
  7. Tag the release with git tag v0.0.0 -m "Release v0.0.0" (0.0.0 replaced with the real ver number).
  8. Build binaries using the Makefile, upload release to Github including the changelog and binaries.

Once its built and released, you need to setup the new development version. To do so:

  1. In irc/constants.go, update the version number to 0.0.1-unreleased, where 0.0.1 is the previous release number with the minor field incremented by one (for instance, 0.9.2 -> 0.9.3-unreleased).
  2. At the top of the changelog, paste a new section with the content below.
  3. Commit the new version number and changelog with the message "Setup v0.0.1-unreleased devel ver".

Unreleased changelog content

## Unreleased
New release of Oragono!

### Config Changes

### Security

### Added

### Changed

### Removed

### Fixed

Updating vendor/

The vendor/ directory holds our dependencies. When we import new repos, we need to update this folder to contain these new deps. This is something that Ill mostly be handling.

To update this folder:

  1. Install https://github.com/golang/dep
  2. cd to Oragono folder
  3. dep ensure -update
  4. cd vendor
  5. Commit the changes with the message "Updated packages"
  6. cd ..
  7. Commit the result with the message "vendor: Updated submodules"

This will make sure things stay nice and up-to-date for users.

Fuzzing and Testing

Fuzzing can be useful. We dont have testing done inside the IRCd itself, but this fuzzer Ive written works alright and has helped shake out various bugs: irc_fuzz.py.

In addition, Ive got the beginnings of a stress-tester here which is useful: https://github.com/DanielOaks/irc-stress-test

As well, theres a decent set of tests here, which I like to run Oragono through now and then: https://github.com/DanielOaks/irctest

Debugging Hangs

To debug a hang, the best thing to do is to get a stack trace. Gos nice, and you can do so by running this:

$ kill -ABRT <procid>

This will kill Oragono and print out a stack trace for you to take a look at.

Concurrency design

Oragono involves a fair amount of shared state. Here are some of the main points:

  1. Each client has a separate goroutine that listens for incoming messages and synchronously processes them.
  2. All sends to clients are asynchronous; client.Send appends the message to a queue, which is then processed on a separate goroutine. It is always safe to call client.Send.
  3. The server has a few of its own goroutines, for listening on sockets and handing off new client connections to their dedicated goroutines.
  4. A few tasks are done asynchronously in ad-hoc goroutines.

In consequence, there is a lot of state (in particular, server and channel state) that can be read and written from multiple goroutines. This state is protected with mutexes. To avoid deadlocks, mutexes are arranged in “tiers”; while holding a mutex of one tier, youre only allowed to acquire mutexes of a strictly higher tier. The tiers are:

  1. Tier 1 mutexes: these are the “innermost” mutexes. They typically protect getters and setters on objects, or invariants that are local to the state of a single object. Example: Channel.stateMutex.
  2. Tier 2 mutexes: these protect some invariants of their own, but also need to access fields on other objects that themselves require synchronization. Example: ChannelManager.RWMutex.
  3. Tier 3 mutexes: these protect macroscopic operations, where it doesnt make sense for more than one to occur concurrently. Example; Server.rehashMutex, which prevents rehashes from overlapping.

There are some mutexes that are “tier 0”: anything in a subpackage of irc (e.g., irc/logger or irc/connection_limits) shouldnt acquire mutexes defined in irc.

We are using buntdb for persistence; a buntdb.DB has an RWMutex inside it, with read-write transactions getting the Lock() and read-only transactions getting the RLock(). This mutex is considered tier 1. However, its shared globally across all consumers, so if possible you should avoid acquiring it while holding ordinary application-level mutexes.

Command handlers and ResponseBuffer

We support a lot of IRCv3 specs. Pretty much all of them, in fact. And a lot of proposed/draft ones. One of the draft specifications that we support is called “labeled responses”.

With labeled responses, when a client sends a label along with their command, they are assured that they will receive the response messages with that same label.

For example, if the client sends this to the server:

@label=pQraCjj82e PRIVMSG #channel :hi!

They will expect to receive this (with echo-message also enabled):

@label=pQraCjj82e :nick!user@host PRIVMSG #channel :hi!

They receive the response with the same label, so they can match the sent command to the received response. They can also do the same with any other command.

In order to allow this, in command handlers we dont send responses directly back to the user. Instead, we buffer the responses in an object called a ResponseBuffer. When the command handler returns, the contents of the ResponseBuffer is sent to the user with the appropriate label (and batches, if theyre required).

Basically, if youre in a command handler and youre sending a response back to the requesting client, use rb.Add* instead of client.Send*. Doing this makes sure the labeled responses feature above works as expected. The handling around PRIVMSG/NOTICE/TAGMSG is strange, so simply defer to irctests judgement about whether thats correct for the most part.