thiell
thiell
We need per-`Worker` `MsgTree` buffering to get more flexibility. This should be implemented using EventHandler evolution.
For now in `Channel` (an `EventHandler`), we are using a `XMLGenerator` over the channel worker object (which acts as a file object). The worker is either `WorkerSsh` for the **parent**...
Need to rework this method in 1.8 (it is a too big change for 1.7.x). Indeed in some rare cases _terminate() could be first called with kill=False and then kill=True,...
Currently ClusterShell only supports reading lines of text. Remove this limitation and add support for binary read/writes.
When using the library, it would be useful/safer to pass the command as list of arguments - as an alternative to string with `shell=True` - as Popen() already does. Looks...
Migrate to **argparse** which is more powerful and would provide an easy way to create tools based on the clustershell library. Will also help for Python 3 support (#239). Usefulness:...
The [Engine](https://github.com/cea-hpc/clustershell/blob/master/lib/ClusterShell/Engine/Engine.py) is a core component of ClusterShell. It is very important to keep it very efficient as it may deal with thousands of clients (EngineClient objects bound to file...
Move fd_max handling to the core library instead of clush only, so that library users and gateways in tree mode can benefit from it.
ev_pickup in tree mode in 1.8 could need some improvements, right now it is blindly called after ev_start(). But it might be difficult to detect when commands really start. TBD.
`ev_written` events are generated when copy or rcopy are used in tree mode, but should not. This issue needs to be analyzed and fixed. Tests are ready to be enforced...