The procedure "gtestsuite" will run the program "gaspan"
under different conditions. With a successful run you
will see the fits flashing by in a seperate window.
gtestsuite utilizes the C-Program gspcompare.c
whichhas first to be compiled: edit Makefile according
to your system compiler and run make.

Run gtestsuite with the command
   gtestsuite /PATH/gaspan test

The procedure will presently run the analysis of 5 different 
spectra under several conditions. The result of a spectrum 
analysis should be quoted with the statement "results are okay".

Differences in the output with respect to the original
data will produce a statement in the form
"look at xxxxx.diff for differences"
where xxxxx characterizes the present spectrum and the
fit options used.

The "test" option will not produce any further output.
In order to obtain output of the analysis you should
replace the "test" option by one of the following options:
"select n" (1 <= n <= 9), "gauss", "tail", "all" or you
can drop any option and select the procedures and
output features by editing gtestsuite.

You can also run the procedure manually by calling gaspan 
and typing:
  GSP> get opt xxx
  GSP> go
where xxx is one of the option files with the extension 
.sta (do not type the extension!).
In some of the option files there will be questions, just
type return as answers.

You might like to have a look at these options files,
they show you how to setup input for gaspan.

If the results of your machine differ from the test run,
you get a message and you can inspect the difference
in a file with the extension .diff

There might be two sources for a difference:
1) different rounding in output: results should only
   differ in the last digit by one unit. This will
   not produce a *.diff file
2) different results but agreement within the statistical
   error of the data. Such effects can occur from different
   rounding procedures during the calculation and are most
   likely to occur in the more complicated fits with tails.
   You can nothing do about this.  *.diff file will be
   generated if the numbers differ too strongly.

Friedrich Riess
e.mail: friedrich.riess@physik.uni-muenchen.de



