Could easily grab it from http://mxr.mozilla.org/servo/search?string=unsafe\s*{&regexp=1&case=on&find=&findi=&filter=^[^\0]*%24&hitlimit=&tree=servo, but that's updated every day.

It would be handy to say things like "is vertical-lr-blocks.html intermittent?" and have crowbot search for I-intermittent issues with that filename in them.

We should write out the known pings each time the list gets updated, either due to getting a new instruction to ping someone or having completed a ping.

We can learn about current build jobs on our machines by fetching http://build.servo.org/json/slaves?as_text=1. One example shows that sevo-linux-cross3 has been running for more than 2 hours: https://gist.github.com/jdm/ae1be5ef10ef308f9fdc93856a0515e0 . We should query this information every 30 minutes and squawk loudly if any build machine is stuck.

We could have crowbot return a random issue to triage (ie. is this issue still valid?), or just the least recently modified issue in the database.

This will require reading the documentation for the node irc library that we're using, presumably.

This will prevent it being misused in attempts to learn more about how the bot works. We should just silently ignore anything like <user> crowbot: explain.

Both 'help' and 'explain' are not exactly helpful.

I've been receiving pings from #servo in #jsapi because I join that channel first, and crowbot hangs out in both.

They're never intended as github references, in my experience.

Is there any reason to use 0.10.x ? The current version is v6.2.2. It provides better ES6 support along with improved performance and bug fixes.

When you type a whole link into IRC, this happens

you> https://github.com/servo/servo/pull/14707
crowbot> PR #14707: Pipeline documentation comments - https://github.com/servo/servo/pull/14707

But if you just type the number in, you get

you> #14707
crowbot> Issue #14707: Pipeline documentation comments - https://github.com/servo/servo/pull/14707

This is dumb. Crowbot has enough information to tell that it's a PR, based on the URL it put in to IRC (also, the GitHub API has a "pull_request" JSON key, so crowbot really doesn't need to parse the URL at all).

09:15 <crowbot1> add `faster` flag to `./mach test` command : https://api.github.com/repos/servo/servo/issues/9273

That links is not as useful as it could be :)

From https://developer.github.com/v3/issues/:

Note: In the past, pull requests and issues were more closely aligned than they are now. As far as the API is concerned, every pull request is an issue, but not every issue is a pull request.

This endpoint may also return pull requests in the response. If an issue is a pull request, the object will include a pull_request key.

eg.: [15:36:01] <pcwalton> anyone: https://github.com/mozilla/servo/pull/1574

TRUMPET THAT FACT

We often notice things that should be noted in TWiS but forget to include them when the time comes. Let's add a command that remembers a phrase and who said it, another command that spits them all out, and a third command that clears the existing set of notes.

https://platform-status.mozilla.org/

The bot gets restarted by heroku at least every 24 hours, so we sometimes drop pending messages. We should write a file of pending pings and read it in at startup.

If I want to talk to someone not in the channel (!summon nick), it would be nice to be notified when that person arrives.

There have been some regressions recently because crowbot has grown organically and it's hard to test everything. We should refactor the bot to separate the interaction with the IRC library from the actual handlers, which would allow us to write automated tests that verify that everything behaves as expected. I imagine the handler function could accept a say argument which would be a function to be called in place of bot.say and bot.act. For the tests this could just append to an array, while for the real bot it would redirect to the appropriate methods of bot.

Similar to 459b4a6, sometimes we want to perform some CI maintenance when there are no active build jobs. We should add a command that pings a user the next time that the queue becomes idle (and check the queue more frequently than once an hour when someone is waiting to be pinged).

It currently responds with a link to the Servo documentation, but I've never seen the command used for that purpose. Users are more likely to want information about crowbot's features, so we should provide that (probably by linking to a document in this repository).

cc @carlosmcgregor

We're seeing notifications for build jobs which have only started <10 minutes ago, so it appears the ETA cannot be trusted at all. We should instead look at the start time for the running build and notify the channel if it has been running for more than 90 minutes. Additionally, we should ensure that our time comparisons take place in the PST timezone, since there's no guarantee that the current time for the bot matches the time for the build log.

At minimum we could have a hash of the most common ones: http://logs.glob.uno/?a=search&c=mozilla%23servo&q=w3.org%2FTR&ss=24+Mar+2013&se=24+Feb+2015

rust-internals could make use of this bot, but it would need to fetch meetings notes and issues from a different place.

It would be nice if crowbot responded to pms to allow people to experiment without cluttering the main channel.

09:13 <jdm> crowbot: what prs need a reviewer?
09:13 <crowbot> add `faster` flag to `./mach test` command : https://api.github.com/repos/servo/servo/issues/9273
09:13 <crowbot> Support device-pixel-ratio in wpt reftests.: https://api.github.com/repos/servo/servo/issues/9272
09:13 <crowbot> Add build-essential to Debian based Linuxes' requirements: https://api.github.com/repos/servo/servo/issues/9264
09:13 <crowbot> mozbrowsersecuritychange event: https://api.github.com/repos/servo/servo/issues/9244
09:13 belew.mozilla.org has kicked crowbot (Channel flood triggered (limit is 5 lines in 3 secs))

We should do this on a timer so it's one every second or something.

Specifically, I'd like to say something like "crowbot: what issue should I poke?" and get back a random issue that:

• either has the E-easy or E-less easy label applied
• has the C-assigned label applied
• has not been updated in at least 14 days
  This should be possible to do using the existing findIssue function by passing the appropriate github API arguments (see the command for reviewing a PR for an example).

The warning added in #75 is no longer always correct since we started using taskcluster.

Sometimes it's useful to know what time it is in London/Seoul.

We're using an old version of node that doesn't like some of the newer syntax in the newsflash module.

It's pretty annoying.

Crowbot currently will turn issue references like #123 (and £123) into links to the main Servo repo. It would be nice if this also worked for other repositories since we use a bunch of repos, with the syntax in the issue title.

We use the assignee for a PR in servo/servo to indicate who is responsible for reviewing the changes. We should add a command that finds all of the PRs with no assignee and reports the title and url, one per line.