travis-cargo no longer works with coveralls, and appears
to be unmaintained. This change use regular cargo to
build and test. It now uses cargo-travis to push test
results to coveralls. This will also build nightly, and
only push test results on the stable build.
In particular, the API for `IdleHandle` now reflects that it is only
really meant for single-use. It mutably borrows the `Client`, so once
`wait` returns there isn't really a good reason to keep the `IdleHandle`
around (because you'll likely want to issue some other commands).
There is something to be said for being able to operate on the IDLE
stream, but we'll leave that for later.
This also avoids some unfortunate unavoidable panics when the connection
fails while the client is IDLEing.
* Use docs.rs for documentation
This would also allow deleting the `gh-pages` branch (which is pretty large).
I'd also recommend adding `badges` and `categories` to make the crates.io page more useful (see [this](http://www.integer32.com/2017/01/20/categories-and-ci-badges.html)), but I'll leave that up to you.
* Remove more references to old docs
* Use bufstream for better read/write
* Read with length 0 == EOF
* Adapt read_delay test to write one chat at a time
* Add test for eof reads
* Neater interface for MockStream
* Add IMAP IDLE support
This patch adds support for the IMAP IDLE command specified in RFC 2177.
It allows clients to block while waiting for changes to a selected
mailbox, without having to poll the server. This is especially useful
for monitoring a mailbox for new messages.
The API is currently somewhat primitive: users can call `Client::idle()`
to get an IDLE "handle", which they can then call `wait()` on to block
until a change is detected. The implementation also provides
`wait_keepalive()` (which tries to avoid connection timeouts) and
`wait_timeout()` (which allows the user to specify the maximum time to
block).
Further down the line, the handle should probably also implement
`Iterator` to allow clients to watch new events as they happen. This
*could* be implemented today, but given that most other `Client` methods
just return unparsed strings at the moment, I didn't feel like that
enhancement was a high priority. For now, users will have to manually
query the mailbox for what changed.
Fixes#26.
* Avoid unnecessary as_bytes()
* Make wait_keepalive interval configurable
* Avoid ?, which requires Rust 1.13
