At the current time, we lack the capacity to test Macintosh submissions, but are fully equipped to run many PC-based programs under Windows 9x/NT/2K. The EML is also capable of testing Unix programs under Solaris 7. Software platforms include, but are not limited to, TSP, SAS, S, SPLUS, STATA, RATS, GAUSS, MATLAB, MAPLE, SYSTAT, SHAZAM, OX, BLSS. Compilers include Fortran, C, C++, Perl. We also support the NAG library.
The instructions must include a ``table of contents'' describing all of the program and data files that are included in the archive entry. Directory structures should also be explained.
Identify the platform on which the executable(s) will run if your submission is platform-specific. For example, "Unix, Solaris X" or "PC, Windows 9x/NT/".
If a submission is comprised of a series of executables, it should be clearly noted in what order they are used.
If stderr is redirected by a Unix-based program, it should be explicitly stated up front.
All submissions are assumed to be batch programs, unless explicitly described as being interactive.
If a Unix-based program spawns additional shells, please add a note to that effect and flag potential problems, if known.
Programs that make heavy use of RAM should have a descriptive sentence about the amount of RAM expected, whether the program uses core RAM or pages to virtual memory, and how much swap is needed. For example, it may not be known to everyone who uses TSP that TSP version 4.x requires a minimum of 4M RAM. Similarly, the Unix version of GAUSS that runs on Solaris 7 defaults to 10M RAM, but many GAUSS programs actually require more.
If a program creates temporary working files/directories, does it occur in the home directory, a specific other directory, or a default swap space (such as /tmp on Unix systems). If a program terminates abnormally, describe whether temporary files are erased automatically.
If the program has built in traps or has been written to ignore normal interrupts, please provide a warning.
In general, we hope that your description of the installation process will omit the need for guess-work and that there will be no surprises waiting when the test problem is finally run.
Grace Katagiri
Katagiri@econ.Berkeley.EDULast modification: 25 April 2001