Author: Greg Egan
(minor updates by Jay Pantone)
Date: 27 May 2019
NB: These notes include some features that are only present from version 13.2 onwards. Please always install the latest version of the program.
A Java client is available in the repository. This consists of the file
DCM.jar , and it requires Java release 8, which can be downloaded
DCM.jar can be run simply by double-clicking it on the desktop for most operating systems, or by typing
java -jar DCM.jar into a
command line (assuming the jar file has been placed in the current directory).
Once you run the Java client, simply type in your team name and hit the
Register button. You can quit the program in either of two ways:
Finish task & quit, which continues with the current task until it is completed, or
Give up task & quit which stops work on the task
immediately and tells the server to assign it to someone else, before quitting.
The Java client does not read or write any files at all. A log of its interactions with the server is available in a scrolling text box at the top of the application's window.
Depending on your operating system and security settings, you might receive warnings against running a file downloaded from the internet.
To override these, you might need to open the jar file the first time by right-clicking it once and choosing
Open, rather than double-clicking.
If you have any concerns yourself about the security of the program, you can verify that it will run in "sandbox" mode by typing the following into a command line (assuming the jar file has been placed in the current directory):
jar xf DCM.jar META-INF/MANIFEST.MF
This will extract the manifest file from the jar and place it in the file
META-INF/MANIFEST.MF. If you examine this file with a text editor,
you can check that the
Permissions property listed in the manifest is
sandbox, which means the Java application will be run without
the permission to read or write any files on your computer.
Windows ready-to-run client and scripts
A precompiled client for 64-bit versions of Windows is available in the Windows64 folder of the repository. Follow the instructions in the
README file inside that folder to install and use this client, along with scripts that automate the process of starting and stopping or quitting
Building and testing the C client
DistributedChaffinMethod.c is a single, standalone file for a command-line C program. It is intended to compile and run under MacOS, Linux and some versions of Windows.
In order to run correctly, the program needs:
- Write access to the current directory.
- Permission to make outgoing connections to the internet.
- The presence of the "curl" command line tool, and the ability for the program to run it via the system() call in the C standard library. These are standard in MacOS/Linux, but for Windows will depend on your precise environment.
Note that the program uses functions in
math.h, so with some compilers it will require the switch
-lm to link with the mathematical
If the program compiles correctly, running it with the option "test" will simply test whether or not it can connect to the server and read back the expected response:
$ DistributedChaffinMethod test Random seed is: 1557505343 Fri May 10 07:39:29 2019 Program instance number: 1314126518 Fri May 10 07:39:29 2019 Team name: anonymous Fri May 10 07:39:29 2019 To server: action=hello Fri May 10 07:39:29 2019 Server: Hello world.
If you wish to commit to running the program to assist in the distributed search, simply run it with no arguments and it will loop indefinitely, waiting for available tasks to execute. Note that:
- To make an effective contribution, the program needs to have (more or less) uninterrupted access to the internet; if it can't make a connection, it will loop, sleeping for a few minutes then trying again, rather than performing any useful computations.
- If a task completes in less than a certain amount of time, your program might sleep for the remainder of the time, to avoid bombarding the server with too much traffic.
- If you wish to quit the program, there are several choices, discussed in the section Shutting down the program. It is much better if you can follow one of these methods, rather than killing the program while it is working on a task.
You can monitor the ongoing results of the search at:
Building under Windows
There are a wide variety of C compilers, IDEs and runtime environments for Windows, and it is impossible for us to test them all in advance, but we encourage Windows users to first try to compile and run WindowsBugTest.c, which can be found in the Debugging folder of the main repository.
If this program successfully compiles (which might need some switches set to ensure compatibility with modern versions of C) then running it will attempt to perform a very short computation, without making any contact with the server. If the output you see ends like this (apart from the specific dates, of course):
Thu May 9 13:07:33 2019 Assigned new task (id=0, access=0, n=6, w=105, prefix=1234561234516234512634512364512346512436512463512465312465132465134265134625134652134562135, perm_to_exceed=567, prev_perm_ruled_out=569) Thu May 9 13:07:33 2019 Finished current search, bestSeenP=490, nodes visited=13910255
then the computation was completed successfully, and you can go ahead and build the full program.
However, if the calculation is not completed, and WindowsBugTest crashes before printing the final line shown above, you will need to find an alternative set of C tools in order to proceed.
Shutting down the program
It is best if the program can be left running continuously, but if you do kill the program or shut down the computer it is running on, then (eventually) the task it was assigned will be reassigned to another instance of the program.
You can tell the program to shut down cleanly, after it has finished with the current task, by placing a file in its working directory with the name:
where NNNNNNN is the same program instance number as used to distinguish the log files. Or, if you use the name:
then any instance of the program using that directory will treat that as a signal to shut down. Note that a task might take more than an hour to complete, so the signal from a STOP file might not come into effect for some time.
You can also tell the program to give up on its current task, by placing a file in its working directory with the name:
When it sees one of these QUIT files, the program will tell the server to reassign its task to another client, and then quit. This should happen in less than a minute.
If you are running under MacOS/Linux, you can also:
- Type CTRL-C once to tell the program to quit after it has finished with the current task.
- Type CTRL-C between three and six times, to tell the program to give up on the current task and quit.
- If the program is unable to make contact with the server at all, hitting CTRL-C repeatedly will eventually force it to quit.
Under Windows, CTRL-C will kill the program immediately, so we would prefer that you shut it down by creating a STOP or QUIT file. Any text editor can be used to create a file with the required name, and it doesn't matter what text the file contains. Just remember to delete the STOP/QUIT file if you want to start running the program again (e.g. after an upgrade).
You can specify a team name when launching the program. This team name will show up on the results page next to any witness strings or superpermutations found by the client. The results page also keeps a log of how many tasks each team has completed. To specify a team name, use the "team" argument, followed by an alphanumeric string with at most 32 characters. Spaces are allowed if placed in quotes.
The default team name is "anonymous" if none is specified.
DistributedChaffinMethod team permutators
DistributedChaffinMethod team "spaces work"
You can tell the program to run for a specified number of minutes, after which it will wait to finish the current task before quitting.
The time limit is specified on the command line:
DistributedChaffinMethod timeLimit 120
would tell the program to run for 120 minutes and then quit, after it has finished its current task.
You can also enforce a stricter time limit, using the option
timeLimitHard. If you specify:
DistributedChaffinMethod timeLimitHard 120
then the program will run for 120 minutes and quit, within about 5 minutes of that quota, even if it is in the middle of a task.
The "timeLimit" and "team" arguments can be used at the same time, in either order.
DistributedChaffinMethod timeLimit 120 team "golden eagles"
The program writes files:
- A cumulative log file, "DCMLog_NNNNNNNNNN.txt"
- A temporary file, "DCMServerResponse_NNNNNNNNNN.txt"
where NNNNNNNNNN is a random integer chosen by each instance of the program.