This is part of the overall strategy I would like to take when approaching
issue #11165. The only two I/O objects that reasonably want to be "split" are
the network stream objects. Everything else can be "split" by just creating
another version.
The initial idea I had was the literally split the object into a reader and a
writer half, but that would just introduce lots of clutter with extra interfaces
that were a little unnnecssary, or it would return a ~Reader and a ~Writer which
means you couldn't access things like the remote peer name or local socket name.
The solution I found to be nicer was to just clone the stream itself. The clone
is just a clone of the handle, nothing fancy going on at the kernel level.
Conceptually I found this very easy to wrap my head around (everything else
supports clone()), and it solved the "split" problem at the same time.
The cloning support is pretty specific per platform/lib combination:
* native/win32 - uses some specific WSA apis to clone the SOCKET handle
* native/unix - uses dup() to get another file descriptor
* green/all - This is where things get interesting. When we support full clones
of a handle, this implies that we're allowing simultaneous writes
and reads to happen. It turns out that libuv doesn't support two
simultaneous reads or writes of the same object. It does support
*one* read and *one* write at the same time, however. Some extra
infrastructure was added to just block concurrent writers/readers
until the previous read/write operation was completed.
I've added tests to the tcp/unix modules to make sure that this functionality is
supported everywhere.
Let's try this again.
This is an implementation of mutexes which I believe is free from undefined behavior of OS mutexes (the pitfall of the previous implementation).
This implementation is not ideal. There's a yield-loop spot, and it's not particularly fair with respect to lockers who steal without going through the normal code paths. That being said, I believe that this is a correct implementation which is a stepping stone to move from.
I haven't done rigorous benchmarking of this mutex, but preliminary results show that it's about 25% slower in the uncontended case on linux (same runtime on OSX), and it's actually faster than a pthreads mutex on high contention (again, not rigorous benchmarking, I just saw these numbers come up).
This allows for easier static initialization of a pthread mutex, although the
windows mutexes still sadly suffer.
Note that this commit removes the clone() method from a mutex because it no
longer makes sense for pthreads mutexes. This also removes the Once type for
now, but it'll get added back shortly.
With the recently added double word CAS functionality on 32-bit ARM (enabled via
a 64-bit atomic instruction in LLVM IR), without some extra features enabled
LLVM lowers code to function calls which emulate atomic instructions. With the
v7 feature enabled, proper 64-bit CAS instructions are used on 32-bit arm.
I've been told that v7 for arm is what we should have been doing anyway. This is
overridable by providing some other non-empty feature string.
Turns out this was a little more far-reaching than I thought it was.
The first commit is the crux of this stack of commits. The `io::io_error` condition is completely removed and the `read` and `write` methods are altered to return `IoResult<T>`. This turned out to be an incredibly far-reaching change!
Overall, I'm very happy with how this turned out (in addition with the `unused_must_use` lint). I had to almost rewrite the pretty printer in `libsyntax` as well as the the formatting in `librustdoc` (as one would expect). These two modules do *tons* of I/O, and I believe that it's definitely improved.
This pull request also introduces the `if_ok!()` macro for returning-early from something that returns a result. I made quite liberal use of this in mostly the pretty printer and html renderer, and I found its usage generally quite pleasant and convenient to have. I didn't really feel like adding any other macro while I was using it, and I figured that pretty printing could be nicer, but it's nowhere near horrid today.
This may be a controversial issue closing, but I'm going to say it.
Closes#6163
* All I/O now returns IoResult<T> = Result<T, IoError>
* All formatting traits now return fmt::Result = IoResult<()>
* The if_ok!() macro was added to libstd
Make the definition of epoll_event use natural alignment on all
architectures except x86_64.
Before this commit, the struct was always 12 bytes big, which works okay
on x86 and x86_64 but not on ARM and MIPS, where it should be 16 bytes
big with the `data` field aligned on an 8 byte boundary.
