Providing plug-in support:
Typically, a test program is written to perform certain basic tests on a test
machine. For example, one might write an "obmc_boot" program that performs
various boot tests on the Open BMC machine.
Experience has shown that over time, additional testing needs often arise.
Examples of such additional testing needs might include:
- Data base logging of results
- Performance measurements
- Memory leak analysis
- Hardware verification
- Error log (sels) analysis
- SOL_console
The developer could add additional parms to obmc_boot and likewise add
supporting code in obmc_boot each time a need arises. Users would employ
these new functions as follows:
obmc_boot --perf=1 --mem_leak=1 --db_logging=1 --db_userid=xxxx
However, another option would be to add general-purpose plug-in support to
obmc_boot. This would allow the user to indicate to obmc_boot which plug-in
packages it ought to run. Such plug-in packages could be written in any
langauge whatsoever: Robot, python, bash, perl, C++.
An example call to obmc_boot would then look something like this:
obmc_boot --plug_in_dir_paths="Perf:Mem_leak:DB_logging"
Now all the obmc_boot developer needs to do is call the plug-in processing
module (process_plug_in_packages.py) at various call points which are agreed
upon by the obmc_boot developer and the plug-in developers. Example call
points which can be implemented are:
setup - Called at the start of obmc_boot
pre_boot - Called before each boot test initiated by obmc_boot
post_boot - Called after each boot test initiated by obmc_boot
cleanup - Called at the end of obmc_boot
This allows the choice of options to be passed as data to obmc_boot. The
advantages of this approach are:
- Much less maintenance of the original test program (obmc_boot).
- Since plug-ins are separate from the main test program, users are free to
have plug-ins that suit their environments. One user may wish to log results
to a database that is of no interest to the rest of the world. Such a plug-in
can be written and need never be pushed to gerrit/github.
- One can even write temporary plug-ins designed just to collect data or stop
when a particular defect occurs.
In our current environment, the concept has proven exceedingly useful. We
have over 40 permanent plug-ins and in our temp plug-in directory, we still
have over 80 plug-ins.
Change-Id: Iee0ea950cffaef202d56da4dae7c044b6366a59c
Signed-off-by: Michael Walsh <micwalsh@us.ibm.com>
diff --git a/lib/gen_valid.py b/lib/gen_valid.py
new file mode 100755
index 0000000..1a52ace
--- /dev/null
+++ b/lib/gen_valid.py
@@ -0,0 +1,104 @@
+#!/usr/bin/env python
+
+r"""
+This module provides valuable argument processing functions like
+gen_get_options and sprint_args.
+"""
+
+import sys
+
+import gen_print as gp
+
+
+
+###############################################################################
+def valid_value(var_value,
+ invalid_values=[""],
+ valid_values=[]):
+
+ r"""
+ Return True if var_value is a valid value. Otherwise, return False and
+ print an error message to stderr.
+
+ Description of arguments:
+ var_value The value being validated.
+ invalid_values A list of invalid values. If var_value is
+ equal to any of these, it is invalid.
+ Note that if you specify anything for
+ invalid_values (below), the valid_values
+ list is not even processed.
+ valid_values A list of invalid values. var_value must
+ be equal to one of these values to be
+ considered valid.
+ """
+
+ len_valid_values = len(valid_values)
+ len_invalid_values = len(invalid_values)
+ if len_valid_values > 0 and len_invalid_values > 0:
+ gp.print_error_report("Programmer error - You must provide either an" +
+ " invalid_values list or a valid_values" +
+ " list but NOT both.")
+ return False
+
+ if len_valid_values > 0:
+ # Processing the valid_values list.
+ if var_value in valid_values:
+ return True
+ var_name = gp.get_arg_name(0, 1, 2)
+ gp.print_error_report("The following variable has an invalid" +
+ " value:\n" +
+ gp.sprint_varx(var_name, var_value) +
+ "\nIt must be one of the following values:\n" +
+ gp.sprint_varx("valid_values", valid_values))
+ return False
+
+ if len_invalid_values == 0:
+ gp.print_error_report("Programmer error - You must provide either an" +
+ " invalid_values list or a valid_values" +
+ " list. Both are empty.")
+ return False
+
+ # Assertion: We have an invalid_values list. Processing it now.
+ if var_value not in invalid_values:
+ return True
+
+ var_name = gp.get_arg_name(0, 1, 2)
+ gp.print_error_report("The following variable has an invalid value:\n" +
+ gp.sprint_varx(var_name, var_value) + "\nIt must" +
+ " NOT be one of the following values:\n" +
+ gp.sprint_varx("invalid_values", invalid_values))
+ return False
+
+###############################################################################
+
+
+###############################################################################
+def valid_integer(var_value):
+
+ r"""
+ Return True if var_value is a valid integer. Otherwise, return False and
+ print an error message to stderr.
+
+ Description of arguments:
+ var_value The value being validated.
+ """
+
+ # This currently allows floats which is not good.
+
+ try:
+ if type(int(var_value)) is int:
+ return True
+ except ValueError:
+ pass
+
+ # If we get to this point, the validation has failed.
+
+ var_name = gp.get_arg_name(0, 1, 2)
+ gp.print_varx("var_name", var_name)
+
+ gp.print_error_report("Invalid integer value:\n" +
+ gp.sprint_varx(var_name, var_value))
+
+ return False
+
+###############################################################################