Remove SocketData and now only pass the dart port to epoll.

Remove SocketData and now only pass the dart port to epoll.

This avoids keeping extra data alive.

BUG=

[Søren Gjesse, 2014-02-20 08:15:54]

LGTM!

The description should probably be updated to say this is Linux only, and that
it also removed the pipe from Dart isolates to the event handler.

We need to be careful with the threading now that the event handler methods can
be called from different threads.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
File runtime/bin/eventhandler_linux.cc (right):

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:38: static void AddToEpollInstance(intptr_t
epoll_fd_, int fd, Dart_Port port,
Move all arguments to the secone line for better readability.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:64: : shutdown_(false) {
Does this fit on the previous line?

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:80: event.data.u64 = timer_fd_ | TIMER_BIT;
Can't we just use ILLEGAL_PORT (value 0) here, and avoid using the high bit?

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:194: void
EventHandlerImplementation::SendData(intptr_t id,
This method is no longer sending data, but just making direct calls. We should
rename it to match that behavior.

Maybe we should also add the thread ID asserts that we added to the Mac
directory watcher code to test/document which methods are called in which
thread.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/socket_linux.cc
File runtime/bin/socket_linux.cc (right):

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/socket_linux...
runtime/bin/socket_linux.cc:64: addr.ss.ss_family, SOCK_STREAM | SOCK_NONBLOCK |
SOCK_CLOEXEC, 0));
Does the Mac/Android system calls also support these additional flags?

[Anders Johnsen, 2014-02-20 08:56:21]

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
File runtime/bin/eventhandler_linux.cc (right):

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:38: static void AddToEpollInstance(intptr_t
epoll_fd_, int fd, Dart_Port port,
On 2014/02/20 08:15:54, Søren Gjesse wrote:
> Move all arguments to the secone line for better readability.

Done.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:64: : shutdown_(false) {
On 2014/02/20 08:15:54, Søren Gjesse wrote:
> Does this fit on the previous line?

Done.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:80: event.data.u64 = timer_fd_ | TIMER_BIT;
On 2014/02/20 08:15:54, Søren Gjesse wrote:
> Can't we just use ILLEGAL_PORT (value 0) here, and avoid using the high bit?

Done.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/eventhandler...
runtime/bin/eventhandler_linux.cc:194: void
EventHandlerImplementation::SendData(intptr_t id,
On 2014/02/20 08:15:54, Søren Gjesse wrote:
> This method is no longer sending data, but just making direct calls. We should
> rename it to match that behavior.
> 
> Maybe we should also add the thread ID asserts that we added to the Mac
> directory watcher code to test/document which methods are called in which
> thread.

Done.

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/socket_linux.cc
File runtime/bin/socket_linux.cc (right):

https://codereview.chromium.org/171503009/diff/50001/runtime/bin/socket_linux...
runtime/bin/socket_linux.cc:64: addr.ss.ss_family, SOCK_STREAM | SOCK_NONBLOCK |
SOCK_CLOEXEC, 0));
On 2014/02/20 08:15:54, Søren Gjesse wrote:
> Does the Mac/Android system calls also support these additional flags?

Not sure. Will do follow-up for other platforms.