Here are circuits used for regressions.  Some of these will also run
on Spice and can be used for comparison.  Most are designed to test
partcular features in hopes of clearly identifying the defective
module.  Some are more general.

You can run any ".ckt" file in batch mode by typing "gnucap -b name" from
the shell.  You can run them interactively by starting gnucap then "get
name".  Then play, change probes, rerun, change values, etc.

The file "test" is a script to automate the testing.  It takes 4
arguments, which all must be supplied.  The diagnostics are poor to
nonexistent.

The arguments are:

$0 "./test"  (its name)

$1 the path name of the executable version of gnucap being tested.

$2 the prefix of all .ckt files to test.  Probably, you want "", which
tests everything.  Otherwise, "d_mos" will test anything beginning
with "d_mos".  It does not use the usual globbing, although it should.

$3 The directory to use for the results files.  Running the script
makes an output file for every input file.  It will make the directory
if it doesn't already exist.  It will overwrite files in that
directory without warning.

$4 The directory to use for comparison.  Usually this will be "==".
It contains all of the reference results files.

Example:

./test ../src/O/gnucap "" 0001 ==

.. places the results in 0001, and compares them to the old ones in "==".

It creates a file ".diffs" in 0001 that contains the differences.
Also, it prints a failure message to the console for any that don't
match.  If you don't get any such messages, but do get a list of all
the files, all pass.

Any differences should be investigated.  It is normal to get a few
minor differences due to differences in machine arithmetic.  Any
results that are significantly different indicate a problem.
