Lu Pan
Lu Pan
By default, failed network requests in ClusterRunner won't raise unless the caller set `error_on_failure` explicitly. This is generally not pythonic as it's implicit instead of explicit.
Registering signal handler for SIGTERM on Windows doesn't work, so we need a way to terminate Windows processes cleanly. Otherwise, it could lead to corrupted git repo if we kill...
It's possible to set file permission on windows to allow only the owner to read & write the file, but `stat.S_IMODE(os.stat(self._filename).st_mode)` would be able to give us the accurate file...
So far, this is only happening on Mac. If the slave is not in master's authorized_keys list, slave is going to be stuck forever asking for password from the user....
- these two should be in two repos really - after separating them, we can put the ClusterRunner SDK on PyPI
e.g. client.folder(folder_id='0').upload_async('/path/to/file'), which returns a promise.