Hi Karen,
On 8/11/11 2:00 PM, Karen Tung wrote:
Hi Drew,
Good work, this will indeed make running tests much easier.
I have a couple comments/suggestions.
You define TEST_MAP in line 48-76 of slim_regression_test.py to
have the list of tests to run. This means if we have new tests,
we need to update this file. Furthermore, as a user, I will have
to remember what are the names of tests available, or I have to read
the code in slim_regression_test.py to find out. Perhaps we
can do something like the following to make it more
user friendly and make future maintenance easier.
Yes, we'll have to update the file, but we have to update tests.nose as
things stand right now ... more on this lower.
- Add an option to the slim_regression_test.py script to list
all the tests available and what "keywords" correspond to which tests.
You can simply run the script with the --help argument:
[mox:tests] > sudo ./slim_regression_test.py --help
Usage: slim_regression_test.py [options] [test[,test]...]
Available tests:
group tests
all: libraries, commands
libraries: boot, common, dataobject, engine, ict, logging,
manifest_input, manifest, target, transfer
commands: ai, dc, installadm, js2ai, system-config, text-install
individual tests
ai: cmd/auto-install/test/
boot: lib/install_boot/test
common: lib/install_common/test
dataobject: lib/install_doc/test
dc: cmd/distro_const/checkpoints/test
engine: lib/install_engine/test
ict: lib/install_ict/test
installadm: cmd/installadm/test
js2ai: cmd/js2ai/modules/test
logging: lib/install_logging_pymod/test
manifest: lib/install_manifest/test
manifest_input: lib/install_manifest_input/test
system-config: cmd/system-config/test, cmd/system-config/profile/test
target: lib/install_target/test
text-install: cmd/text-install/test
transfer: lib/install_transfer/test
Options:
-h, --help show this help message and exit
-c CONFIG, --config=CONFIG
nose configuration file to use
--suppress-results suppress the printing of the results
--hudson=HUDSON_NUM hudson job number to use as a baseline
To get the list of all the tests.
- Instead of using a static list, generate the list dynamically.
Of course, doing this will require that we define a convention on
how tests are defined in the slim_source gate. Perhaps we
can do something like the following.
In each of the test sub-directories,we define a text file with a
certain name. A directory with this file
is considered a test directory, and will get automatically added to
the list of tests to be executed. In this file, we can put information
about the "test suite" in the directory, such as test suite name. The
test
suite name will the short name that people can use when they want to
run a certain test suite.
I wouldn't have any issues with reorganizing some of our gate to make
the test directories adhere to some standards, but I'm not sure we want
to make those changes at this point. If we do, I would do things like:
- move the test directory to the root of the library/command (DC's are
in $SRC/cmd/distro_const/checkpoints/test instead of
$SRC/cmd/distro_const/test)
- combine multiple test directories into one test directory. AI and
system-config have multiple directories of tests
Maybe from there, we could try to automate "discovery" of tests, but I
don't know.
Thanks for looking Karen!
-Drew
Thanks,
--Karen
On 08/11/11 08:53, Drew Fisher wrote:
Good morning!
I was hoping to get a quick code review for
6987307 <http://monaco.us.oracle.com/detail.jsf?cr=6987307> Update
slim_test to allow better granularity of test selection
https://cr.opensolaris.org/action/browse/caiman/drewfish/6987307/webrev/
This impacts none of our packaged code, so I figured it was safe to
send out even though we're in a restricted build phase.
I added a new test script which allows two new things:
- the ability to specify subsets of tests to run rather than running
every single test in the gate
- regression testing against prior Hudson results.
The regression testing isn't the smartest algorithm so take that
testing with a grain of salt. By default, the new script looks at
the latest Hudson install_unit_tests job and compares against that.
If somebody pushes something that breaks 50 tests, that will be used
as the baseline. I added a flag to the script which allows you to
specify which Hudson job you want to compare against should something
like that happen.
The change to tests.nose was to re-include the /lib/install_ict/test
directory. It went missing at some point ...
Thanks!
-Drew
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
