How to set a time limit for program execution

[hels]

Thanks mat123.

The problem occurred when the command was run concurrently with the Java program (under Parallels on an dual core Intel Mac).

I have now moved it to a dual core PC where it takes 140 seconds to run. This is 40 seconds above the wanted time limit of 100 seconds. But acceptable for my application.

The remaining problem is how to access the exit status (0 or 1) from the Java program (in case it finishes before the time limit). ERRORLEVEL is not available for that purpose, since it is set by the FIND command.

[Salmon Trout]

I have now moved it to a dual core PC where it takes 140 seconds to run. This is 40 seconds above the wanted time limit of 100 seconds.

Use the right sleep.exe as I suggested and you can get the 100 seconds that you want.

[hels]

Hi Salmon Trout

The inaccurate delay (about 40%) is not caused by the sleep itself, but by the TASKLIST command being executed inside a loop (100 executions), concurrently with the Java program.

Sleeping a few milliseconds, instead of a second, will only increase this inaccuracy since more TASKLIST calls have to be made. In fact, sleeping 2 seconds 50 times results in a delay closer to 100 seconds. The drawback of this "sleep 2" solution is of course that each execution of the Java program will take at least 2 seconds.

[mat123]

hels
after the tsklist|find line you need to put this
if errorlevel 1 (do whaterver you want)

[Salmon Trout]

I fear you misunderstand me

You are doing this

start java program
loop
     check if java is running; if not exit loop
     increment count; if count=100 exit loop
     sleep 1 second (i.e.1000 milliseconds)
endloop

This takes 140 seconds

Surely if you adjust the sleep to around (100/140) seconds - around 700 milliseconds - then the total time will come to around 100 seconds?

Please explain where my logic is faulty




    
    

[Salmon Trout]

The drawback of this "sleep 2" solution is of course that each execution of the Java program will take at least 2 seconds.

This doesn't make sense. You start the java program once only, don't you?

[hels]

You start the java program once only, don't you?

No, the batch program will be called from another batch program in order to solve about 100 puzzles.

Surely if you adjust the sleep to around (100/140) seconds - around 700 milliseconds - then the total time will come to around 100 seconds? Please explain where my logic is faulty

Your logic is not faulty, but it turns out in practice that the calls of 'tasklist' slow down the whole computation substantially when the Java program is running in parallel. Besides, I would like to have a solution that is machine independent.

after the tsklist|find line you need to put this
if errorlevel 1 (do whaterver you want)

This test refers to the result of 'find', not the result of the Java program execution. I want to do something like

    java Program
    if errorlevel ...        // testing the exit status of Program
   
and also make sure the Java program doesn't run for more than 100 seconds.

[Salmon Trout]

Why don't you include code in the Java program to record the start time and halt after 100 seconds?

[hels]

Why don't you include code in the Java program to record the start time and halt after 100 seconds?

The Java program (or actually Java programs) will be submitted by my students. I will use a UNIX script for testing their solutions (no problem in that: ulimit -t 100), but I also would like to provide a DOS batch program so that they can test their solutions on their own PCs.