|This option sets the X display which xmond connects to. display_name can be a name or numerical network address. The default for both values is the value of the DISPLAY environment variable. If this is not set, then the default for display_name is the local host and the default for display_number is 0. If display_number is positive, the real port number is determined by adding it to the base X server port number, 6000, just as is done for standard X servers. If it is negative, the real port number is the absolute value of display_number. A negative value may be useful when using xmond as a general purpose tcp/ip monitor and the byte stream being monitored does not obey the X protocol.|
|This option sets the port on which xmond listens for client connections. This port is always on the host where xmond is running. Default is 1. The real port number is determined in the same way as for the -server option.|
|This option sets the amount of output initially produced by xmond for each request, reply, event or error received. The values are 0 (off), 1 (names), 2 (main), 3 (full), 4 (hex). Default is 0. The meaning of each value is described in the USER INTERFACE section.|
|-raw||If this option is given, xmond simply acts as a tcp/ip monitor, passing bytes from one side to the other without parsing the stream. By default, it prints the hexadecimal value of each byte passed. This behaviour may be altered by using the -ascii or -noprintraw options.|
|If this option is given, dumps are printed as ascii characters, except for unprintable characters, which are printed as hexadecimal, escaped with a #146;. This is useful with the -raw option.|
|If this option is given with the -raw option, xmond does not print anything.|
|If this option is given, xmond initially prints the size of each block of data received from the clients and server.|
|Print size of traffic only. Equivalent to "-packet_size -raw -noprintraw".|
Xmonui accepts all of the standard X Toolkit command line options. In particular
-display display_name:display_number Indicates where to display the user interface window.
1) To monitor connections made to the local host use the following command line
xmonui | xmond
Connections made to the display localhost:1 will appear on the display localhost:0. The xmonui user interface will appear on the display localhost:0.
When starting up the application that you want to monitor, be sure to set its display correctly. When you normally start up an X application, and if you have not done anything special, it will by default start up on display 0 of your local host. But xmon by default is listening as if it is display 1 of your local host. Assuming your X application is called "myclient" and your local host name is "dolphin" then type the following to start up your application:
myclient -display dolphin:1
Another way to do the same thing is to change the DISPLAY environment variable. If you are using a C shell you can do this:
setenv DISPLAY dolphin:1
If you are not using a C shell, you may have to do something different to change the DISPLAY environment variable. Also, most clients understand the -display option, but there are those that do not. To determine the name of your local host try the hostname command (although it may not be available on all machines).
2) If you are on the host squiggle which has two X servers using the displays squiggle:0 and squiggle:1, and want to monitor connections made to the server running on the display juggler:0, enter the following command line.
xmonui -display squiggle:0 | xmond -server juggler:0 -port 2
Connections now made to the "display" squiggle:2 will appear on the display juggler:0. The xmonui user interface will appear on the display squiggle:0. Monitor output will appear in the window where the command was entered.
3) A good way to view the output of xmon is to pipe it through xless, which allows you to scroll and search through the output interactively. For example:
xmonui | xmond | xless -f
The user interface is divided into four parts: output detail, statistics, selected requests and selected events.
In the output detail section, the amount of detail contained in the output of xmon can be selected. Different levels of detail can be selected for each of the message types: requests, events, errors and replies. The meaning of each level is as follows.
In the selected requests section, selected requests can be monitored at a different level of detail, or can be blocked from transmission. Requests can be selected by clicking on their names in the scrollable list. Clicking again de-selects the request. Selected requests are indicated by an asterisk (*) in the scrollable list.
off No monitor output is produced. names Only the names of the messages are output. main The most interesting fields of the message are output. full All fields of the message are output. hex All fields of the message are output, as well as a hexadecimal dump. The detail setting for errors also applies to the following: setup messages sent at client connection; the end-of-file "message" sent at client shutdown; unknown extended messages; and unexpected replies. If xmon receives a request, reply, event or error which it does not know how to handle, and if the detail setting for errors is "hex", then the message will be dumped in hexadecimal. Similarly, if xmon receives a reply for which it did not send a corresponding request, and the detail setting for errors is "hex", then the reply will be dumped in hexadecimal. Note that synthetic events (events sent by XSendEvent) are monitored in the same way as normal events but are identified as being "SYNTHETIC". Also in this section is the show packet size toggle. If this is turned on, xmon will display the size of each packet received from both clients and servers. The file descriptor of the client or server which sent the packet is also displayed. The first client file descriptor is 4. File descriptors 0, 1 and 2 are used by standard input, standard output and standard error and file descriptor 3 is where xmon listens for new connections. In the statistics section, the counting of requests, events and errors can be controlled as follows. start Enable the taking of statistics. stop Disable the taking of statistics. clear Clear the counts for this message group. Print the name and number of occurrences of each message in this group, excluding messages received zero times. print zero Print the names of messages in this group that have been received zero times.
The detail toggle is of the same form as in the output detail section, but applies only to those requests selected in the left scrollable list.
If the blocking toggle in set to on, all selected requests in the right hand scrollable list are blocked by xmon. They are not forwarded to the server, although they are monitored and counted normally. If the blocking toggle in set to off, all requests are forwarded to the server.
The selected events section is similar to the above section but deals with events received from the server.
Normally xmonui is used as an interactive interface to xmond. However, for some testing procedures it may be better to run xmond by itself initialising it with some standard setup. The interface between xmonui and xmond is made up of simple ascii strings. Pressing buttons on xmonui causes it to write these strings to standard output which are then usually read by xmond. You can just run xmond by itself and type in the strings, or, even better, use a file as input to xmond. There are too many strings to list here, but if you run xmonui by itself, you will see the strings being printed to standard output. Run
xmonui > command.file
to create a file of strings that can be used as input to xmond. For example, a file which will cause xmond to monitor the Bell request and also print the names of all events would contain the lines:
xmond < command.file
will then set up xmond in the same way each time. Note that when running xmond by itself, it does not exit on reading end-of-file and so it must be killed. (I use CONTROL-C to kill. Your kill character may be different.)
It is also possible to initialise xmond with a file, and then take further commands from xmonui by running
xmonui | cat command.file - | xmond
The cat command first sends the command.file to xmond and then passes the output of xmonui to xmond. Please note that in this case xmonui will not show the new state of xmond correctly. For example, xmonui starts up assuming that no requests have been selected, even if the command.file has initialised xmond by selecting some requests. This inconsistency does not effect the usability of xmon but may be confusing.
will cause xmond to exit, so make sure that this does not occur in any input file. Also it is meaningless to use the statistics commands from within an input file because these will be read before any X clients have connected.
Xmon contains an unfinished attempt at event recording and playback. It is fairly crude and does not work properly. If you are interested in exploring it and perhaps doing some more work on it, please try it out. To compile it in, you will need to modify the Imakefile as suggested in the comments.
To try it out, first run xmond with the -record option and with standard output redirected to a file. The -record option causes xmon to write, to standard output, a log of all user events it receives. Run an X client through xmond and interact with it. The mouse and keyboard events will be logged to the file. Now kill that invocation of xmond.
Now run xmond with the -play option and with standard input redirected from the event log file just produced. Now run the same X client as before, but do not interact with it. The client should receive events from xmond according to the log file and behave in the same way as before.
The events are played back at the same speed as when recorded (thanks to code contributed by Marc Vertes). The algorithm used to decide when the client is ready to receive the next event is faulty and can sometimes wait forever. It simply counts the number of ImageText8 and PolyText8 requests that are received and records the counts with each event in the log file. During replay, an event is sent to the client when the matching number of requests have been received.
For example, to record and playback a simple xterm session:xmond -record > foo (then run xterm through it, typing a few commands) (exit xterm and kill xmond) xmond -play < foo (now run xterm through it again)It half works. Good luck.
No provision is included for extensions to the base protocol.
Xmon only handles TCP socket connections. UNIX domain sockets and DECnet are not supported.
There should be a better way of initialising the state of xmond and having this new state reflected in xmonui.
Because of the security method used by your X server, it may reject connection attempts made by xmond. A work-around is to run
to force the X server to accept all connections.
X Window System Protocol, MIT X Consortium Standard, X Version 11, Release 4, by Robert W. Scheifler.
Greg McFarlane, OTC, Australia, from the xscope program written by James L Peterson, MCC.
Please send any bug reports or modifications to
|X Version 11||XMON (1)||30 March 1996|