Basic support for Unix domain sockets on Windows within
the limitations of the operating system itself: no
support for abstract addresses, datagram or
packet sockets and ancillary data.
This also fixes File::GetType on Windows to correctly
identify reparse points representing AF_UNIX sockets
as such, rather then identify them as links.
Finally, it is worth noting that there is an existing
discrepancy between POSIX OSes and Windows: on POSIX OSes
File(sockPath).existsSync() returns true, but on Windows the
same code returns false because Unix domain sockets are not
considered regular files by stat.
This CL does a bunch of refactoring around RawAddr class which
surves as a wrapper around various structures in sockaddr_* family.
Distinguishing anonymous AF_UNIX address from abstract AF_UNIX
address requires passing around the length of the address
structure. Thus we incorporate this information into
RawAddr. This will also make possible to properly support
full-range of abstract AF_UNIX addresses in the future because
supporting these requires properly handling embedded NUL bytes.
(See https://github.com/dart-lang/sdk/issues/46158).
Fixes https://github.com/dart-lang/sdk/issues/41161
TEST=standalone/io/unix_socket
Cq-Include-Trybots: luci.dart.try:vm-fuchsia-release-x64-try
Change-Id: I016cb33ebdd62f0cac1ae97d822105366a6a6964
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/457720
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Android is based on Linux, so most of the files were identical
sans some subtle discrepancies caused by drift over time.
This discrepancies were making code base harder to maintain and in fact
were hiding bugs. For example, on Android eventhandler's implementation
of timers which relied on passing timeout to `epoll_wait` contained
a bug which was not present on Linux which used `timerfd` instead.
TEST=ci and manual testing of Flutter app on Android device
Fixes https://github.com/dart-lang/sdk/issues/54868
Cq-Include-Trybots: luci.dart.try:vm-aot-android-release-arm64c-try,vm-aot-android-release-arm_x64-try,vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64c-try,vm-ffi-android-product-arm-try,vm-ffi-android-product-arm64c-try,vm-ffi-android-release-arm-try,vm-ffi-android-release-arm64c-try
Bug: b/311165013
Change-Id: Ia166f69c14177ec34160805a0983eafee8ea65f6
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/350923
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
We are using `epoll()` in edge-triggerred mode, which requires
us to hit `EAGAIN` on `write()` before we are guaranteed
to receive an `EPOLLOUT` event.
`_SocketStreamConsumer` looks at `_NativeSocket.writeAvailable`
after issuing a `write` to decide whether it should continue
writing or pause and wait until event handler sends us `writeEvent`.
However we did not previously set `writeAvailable` to `true` until
the first `writeEvent` arrived, which lead to a hang on some
Linux servers: we would write a small amount of bytes into a pipe
and then wait for `writeEvent`, which would never come, as `epoll()`
is only guaranteed to wake up with `EPOLLOUT` if we hit `EAGAIN`
on `write()`.
This CL also changes `_NativeSocket.nativeWrite` implementation
to write bytes into the file descriptor until it gets EAGAIN.
This reduces the number of roundtrips between Dart and
native.
Unfortunately this CL does not come with a regression test
because it is relying on concrete `epoll()` behaviour and it this
behaviour does not reproduce for me.
TEST=Tested manually on the affected servers.
FIXED=b/286346121
Fixed: 286346121
Change-Id: I37fef9aa12b1da724b035aa9577b414a8057217e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/308101
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Ivan Ugryumov