binder: Refactor: move FdTrigger to its own file / class.
Also move interruptable*Fully functions to RpcTransport so that we no
longer need pending() and pollSocket().
This also allows us to hide send() / recv(); callers should use
interruptableWriteFully / interruptableReadFully instead, because
those repsect the shutdown trigger.
- Fix one place to use interruptableWriteFully() instead of send() when
sending header.
interruptable*Fully are marked as virtual functions because TLS will
need to poll with events dynamically adjusted. See follow-up CLs for
TLS implementation.
Test: TH
Bug: 190868302
Change-Id: I131eed3a637b3a30280b320966e466bbfac0fc45
diff --git a/libs/binder/RpcState.cpp b/libs/binder/RpcState.cpp
index 23382c3..b58f1b3 100644
--- a/libs/binder/RpcState.cpp
+++ b/libs/binder/RpcState.cpp
@@ -283,8 +283,8 @@
}
if (status_t status =
- session->mShutdownTrigger->interruptableWriteFully(connection->rpcTransport.get(),
- data, size);
+ connection->rpcTransport->interruptableWriteFully(session->mShutdownTrigger.get(),
+ data, size);
status != OK) {
LOG_RPC_DETAIL("Failed to write %s (%zu bytes) on RpcTransport %p, error: %s", what, size,
connection->rpcTransport.get(), statusToString(status).c_str());
@@ -305,8 +305,8 @@
}
if (status_t status =
- session->mShutdownTrigger->interruptableReadFully(connection->rpcTransport.get(),
- data, size);
+ connection->rpcTransport->interruptableReadFully(session->mShutdownTrigger.get(),
+ data, size);
status != OK) {
LOG_RPC_DETAIL("Failed to read %s (%zu bytes) on RpcTransport %p, error: %s", what, size,
connection->rpcTransport.get(), statusToString(status).c_str());