https://github.com/rust-lang/rfcs/pull/221
The current terminology of "task failure" often causes problems when
writing or speaking about code. You often want to talk about the
possibility of an operation that returns a Result "failing", but cannot
because of the ambiguity with task failure. Instead, you have to speak
of "the failing case" or "when the operation does not succeed" or other
circumlocutions.
Likewise, we use a "Failure" header in rustdoc to describe when
operations may fail the task, but it would often be helpful to separate
out a section describing the "Err-producing" case.
We have been steadily moving away from task failure and toward Result as
an error-handling mechanism, so we should optimize our terminology
accordingly: Result-producing functions should be easy to describe.
To update your code, rename any call to `fail!` to `panic!` instead.
Assuming you have not created your own macro named `panic!`, this
will work on UNIX based systems:
grep -lZR 'fail!' . | xargs -0 -l sed -i -e 's/fail!/panic!/g'
You can of course also do this by hand.
[breaking-change]
The tcp-accept-stress, despite the previous modifications, is still deadlocking
on the osx buildbots. When building/testing/running repeatedly locally, it was
discovered that the test would often fail with TcpStream::connect returning the
error `address not available`.
This test opens up quite a large number of sockets, and it looks like by default
osx isn't the speediest at recycling those sockets for further use.
The test has been modified (and verified) to not deadlock in this error case,
and the test is not just officially ignored on OSX (with no FIXME). I believe
that we'll get good coverage of the relevant code on the linux builders, so this
isn't so much of a loss.
At the same time I turned down the stress parameters to hopefully lighten the
socket load on other platforms.
It was previously asserted that each thread received at least one connection,
but this is not guaranteed to always be the case due to scheduling. This test
also deadlocked on failure due to a lingering reference to the sending half of
the channel, so that reference is now also eagerly dropped so the test can fail
properly if something bad happens.
Closes#16872
This commits takes a similar strategy to the previous commit to implement
close_accept and clone for the native win32 pipes implementation.
Closes#15595
This commits implements {Tcp,Unix}Acceptor::{clone,close_accept} methods for
unix. A windows implementation is coming in a later commit.
The clone implementation is based on atomic reference counting (as with all
other clones), and the close_accept implementation is based on selecting on a
self-pipe which signals that a close has been seen.
