- Fix: Changed Preferences to inherit privately from Ui::Preferences
- Added "About Qt" to Help Menu
- Added the PortList/StreamList actions to File Menu
- Fix: "New Stream" action is enabled only when a port is selected in the port list
- Fix: Ostinato now looks for Drone in the correct path
- Win32: If bindconfig.exe is not present, WinPcapPort::hasExclusiveControl() returns false instead of true (we assume we don't have exclusive control)
- Win32: WinPcapPort now looks for bindconfig.exe in the app's dirPath() instead of the current directory
- Top level 'make clean' and 'make distclean' now do not stop in case of errors
- 'make distclean' now removes the object_script.* generated files
- Added the Logo to the About dialog and also the application icon
- The default 'all' target no longer has 'qmake' as prerequisite; 'make qmake' shd be done manually on a pristine codebase
- eol's changed from dos-style to unix-style
- Trivial changes in sample.h/sample.cpp
- Implemented ARP protocol
- Top level Makefile now has a 'release' target
- pcap_set_direction() is no longer invoked in case of Win32 as older versions of WinPcap do not support this API
- MyService::modifyPort() declaration modified to fix the error reported by newer gcc compiler
- Queued RPC calls would cause crashes due to invalid pointers to request/response and/or controller; this has been fixed
- PbRpcController now takes ownership of request and response messages and
will delete them when it itself is being deleted
- This design mandates that request and response messages for each RPC call
have to be allocated on the heap.
- The convention for the Closure 'done' call now is to allocate and pass a
pointer to the controller object to it which will delete it after use;
this requires that controller itself be also allocated on the heap
(NOTE: this is just a convention - not mandatory)
- All existing RPC calls (in portgroup.cpp) have been changed to follow the
above convention
- Reordering of queued RPC calls has been fixed
- PortManager is now destroyed at exit; because of this fix the per port temporary capture files are auto-removed at exit
- WinPcapPort destructor no longer deletes the monitor threads because the parent class PcapPort already does it
- Capture does not automatically (and incorrectly) stop after one packet if started immediately after a View Capture operation
- User is prompted to stop transmit on a port first if he tries to apply configuration changes on a port in 'transmit' state
- OstProto service has a new method "modifyPort()"
- At port init port.isExclusive is now set using a bindconfig query (Win32 only)
- AbstractPort interface has 2 new pure virtual methods - hasExclusiveControl() and setExclusiveControl()
- PcapPort does not support this functionality (yet) so these methods return false
- WinPcapPort suppots this new functionality using bindconfig
- Port's notes (specifying Rx/Tx limitations) are now set and updated based on hasExclusiveControl()
- Presence of 'notes' on a port is indicated using a '*' after the port name in the port stats window
- The tabwidget has been removed from Port Window | Stream View Pane
- Ostinato Client has a new action in the port window's context menu for the same
- Port Icon in the Port Window is decorated based on exclusive control
- Server now updates the "packet list" after every "apply" rather than before a "start transmit" (if dirty) - this better reflects user's expectation from these operations
- Client disables the entire application and changes to a "Busy" cursor when "applying" stream configuration
Fixes
- UDP checksum no longer is zero but a valid value
- Order of streams no longer gets messed up across a "apply"
- (Abs/Pcap/WinPcap)Port: Inter SendQueue Timing implemented using a dummy packet length of zero (this is not supported by the winpcap pcap_sendqueue_transmit() but since we don't use that API we are not affected)
- (Abs/Pcap/WinPcap)Port: Loop Mode Timing across loop iterations implemented
- Added the year 2010 to Copyright notice in the About dialog
Fixes
- (Win)PcapPort: Fixed the wrong rates when packets sent in 'burst' mode
- Major code reorganization of the server code across several classes with fewer 'friends'
- New server classes - AbstractPort, PcapPort, WinPcapPort, PortManager
- With this reorg classes have more focus than earlier and will be hopefully easy to extend
Fixes
- Ostinato client is now able to successfully reconnect and talk to the Ostinato server after a disconnect - earlier, if a method had been pending during the disconnect, the communication was not up after a reconnect; pending methods are cleaned up at disconnect now
- Ostinato Client
- will start the server as a child process at startup and terminate it at exit
- Ostinato Server (Drone)
- is now a system tray application
- if not able to bind to a IP/Port successfully, informs the user and exits
- the GUI is now nothing more than a TextLabel
Others
- If a getStats() request is pending, the client will not queue up any more requests till a reply is received for the pending one
- Nitpicks in the Payload protocol Widget, PortsWindow Widget
- UserScript Protocol reworked. Needs more testing. May need a significant rewrite in the future, but for now this will have to do - therefore it has been marked "EXPERIMENTAL" for now
- LinkState is now updated in PortWindow as soon as a change is detected; the required minimal refactoring of the Port class usage has been done
- Fixed a compiler warning in portgrouplist.cpp
Others
- PortStatsFilter: Ui change - added left and right arrow icons
---------------------------------
- StreamConfigDialog: Valid subsequent protocol choices for a particular protocol in the simple protocol selection widget is no longer hardcoded - ProtocolManager is queried for validitity of each pair of possible protocols; signal-slot connections are made accordingly. This refactoring makes it easier to add a protocol to the simple protocol selection widget
- ProtocolManager: populates and maintains a database of valid 'neighbour protocols' and implements a method - isValidNeighbour() to query the same for a pair of protocols
- AbstractProtocol: new method protocolIdType() introduced to build the above database (in conjunction with the existing method protocolId(ProtocolIdType)); default implementation returns ProtocolIdNone
- Protocols which include a valid/supported ProtocolIdType (eth/llc/ip) reimplement protocolIdType() to return the apporpirate ProtocolIdType. These are viz.
- combo
- eth
- llc
- snap
- ip
- Speed optimization while populating streamqueues if the protocolFrameValue/Size() does not vary across packets
- AbstractProtocol: new methods to support the above optimization
- isProtocolFrameValueVariable()
- isProtocolFrameSizeVariable()
- isProtocolFramePayloadValueVariable()
- isProtocolFramePayloadSizeVariable()
(each of the default implementations returns false indicating that the protocol frame value or frame size is fixed and not variable)
- Protocols which support variable values/size (list follows) reimplement the above methods appropriately
- combo
- mac
- dot3
- ip4
- tcp
- udp
- payload
- StreamInfo::makePacket() moved to base class as StreamBase::frameValue()
- StreamBase: all 'get' accessor functions made 'const'
- class ProtocolManager: while registering a protocol, no need to pass the protocol name; ProtocolManager finds it out internally by using the protocol's shortName() method
Fixes
-----
- Fixed issue with port capture not starting the first time 'start capture' was clicked
- Port State (Link/Transmit/Capture) now updated alongwith port stats
- On link state change, the port window is not updated - partial changes have been done under #if 0; needs refactoring of Port Class implementation/usage before a signal/slot for the same can be implemented
Fixes
- Fixed crash in client when connection to server is broken
- Packet Capture and Capture Buffer Retrieval now works correctly and consistently (I think!)
Others
- Minor visual changes in Ports Window
- Port Stats Window now has 'right' alignment for stats data and 'center' for state data
- Added support for retrieving the packet capture buffer from server to client (does not work consistently however - needs investigation)
- getCaptureBuffer() Rpc signature changed
- RPC: Added support in Rpc Channel (client) to queue calls
- RPC: Added support for transferring arbitrary binary data from server to client (used to get packet capture files)
- Rpc header changed - length is now 4 bytes instead of 2; there is no rsvd field any longer
Fixes
- RPC: Fix for the case when a msg is not received all at once over the socket
- StreamConfigDialog: fixed display issue in packet view for combo protocols containing meta fields
- Fixed issue with Stacked Vlan not retaining data for both CVlan and SVlan
- Fixed incorrect payload size issue with increment/decrement frame length modes
Refactoring, Cleanup etc.
- RPC: Minor code and TODOs cleanup
- Server: Minor code and TODOs cleanup
- Server: Removed unused file(s): rxtx.cpp, rxtx.h
- Server: Replaced direct use of ProtocolList with the ProtocolListIterator
- Common: Minor code and TODOs cleanup
- StreamBase::frameLen() now returns the length based on the mode/min/max and the passed in streamIndex
- AbstractProtocol interface changed for methods - protocolFrameSize(), protocolFrameOffset(), protocolFramePayloadSize() : all of them now take streamIndex as an optional param with 0 as the default value
- Protocols implementing the above methods changed accordingly
