aboutsummaryrefslogtreecommitdiffstats
path: root/test/RPCSession.cpp
diff options
context:
space:
mode:
authorBob Summerwill <bob@summerwill.net>2016-08-01 13:25:37 +0800
committerBob Summerwill <bob@summerwill.net>2016-08-01 16:45:11 +0800
commit4ee2114127f87b08b76b3ca94cde80a49cdc056a (patch)
treeb680926d0da4aadfddae0db9567557802f2c2929 /test/RPCSession.cpp
parent56727d61a61e1485c8360f00700d766632ec7163 (diff)
downloaddexon-solidity-4ee2114127f87b08b76b3ca94cde80a49cdc056a.tar.gz
dexon-solidity-4ee2114127f87b08b76b3ca94cde80a49cdc056a.tar.zst
dexon-solidity-4ee2114127f87b08b76b3ca94cde80a49cdc056a.zip
Make the Solidity repository standalone.
This commit is the culmination of several months of work to decouple Solidity from the webthree-umbrella so that it can be developed in parallel with cpp-ethereum (the Ethereum C++ runtime) and so that even for the Solidity unit-tests there is no hard-dependency onto the C++ runtime. The Tests-over-IPC refactoring was a major step in the same process which was already committed. This commit contains the following changes: - A subset of the CMake functionality in webthree-helpers was extracted and tailored for Solidity into ./cmake. Further cleanup is certainly possible. - A subset of the libdevcore functionality in libweb3core was extracted and tailored for Solidity into ./libdevcore. Further cleanup is certainly possible - The gas price constants in EVMSchedule were orphaned into libevmasm. - Some other refactorings and cleanups were made to sever unnecessary EVM dependencies in the Solidity unit-tests. - TravisCI and Appveyor support was added, covering builds and running of the unit-tests (Linux and macOS only for now) - A bug-fix was made to get the Tests-over-IPC running on macOS. - There are still reliability issues in the unit-tests, which need immediate attention. The Travis build has been flipped to run the unit-tests 5 times, to try to flush these out. - The Emscripten automation which was previously in webthree-umbrella was merged into the TravisCI automation here. - The development ZIP deployment step has been commented out, but we will want to read that ONLY for release branch. Further iteration on these changes will definitely be needed, but I feel these have got to sufficient maturity than holding them back further isn't winning us anything. It is go time :-)
Diffstat (limited to 'test/RPCSession.cpp')
-rw-r--r--test/RPCSession.cpp16
1 files changed, 13 insertions, 3 deletions
diff --git a/test/RPCSession.cpp b/test/RPCSession.cpp
index 54e1675f..e9df9031 100644
--- a/test/RPCSession.cpp
+++ b/test/RPCSession.cpp
@@ -36,15 +36,25 @@ IPCSocket::IPCSocket(string const& _path): m_path(_path)
BOOST_FAIL("Error opening IPC: socket path is too long!");
struct sockaddr_un saun;
+ memset(&saun, 0, sizeof(sockaddr_un));
saun.sun_family = AF_UNIX;
strcpy(saun.sun_path, _path.c_str());
+// http://idletechnology.blogspot.ca/2011/12/unix-domain-sockets-on-osx.html
+//
+// SUN_LEN() might be optimal, but it seemingly affects the portability,
+// with at least Android missing this macro. Just using the sizeof() for
+// structure seemingly works, and would only have the side-effect of
+// sending larger-than-required packets over the socket. Given that this
+// code is only used for unit-tests, that approach seems simpler.
+#if defined(__APPLE__)
+ saun.sun_len = sizeof(struct sockaddr_un);
+#endif // defined(__APPLE__)
+
if ((m_socket = socket(AF_UNIX, SOCK_STREAM, 0)) < 0)
BOOST_FAIL("Error creating IPC socket object");
- int len = sizeof(saun.sun_family) + strlen(saun.sun_path);
-
- if (connect(m_socket, reinterpret_cast<struct sockaddr const*>(&saun), len) < 0)
+ if (connect(m_socket, reinterpret_cast<struct sockaddr const*>(&saun), sizeof(struct sockaddr_un)) < 0)
BOOST_FAIL("Error connecting to IPC socket: " << _path);
m_fp = fdopen(m_socket, "r");