Add some desired options to the au test harness.

Add some desired options to the au test harness.

Adds verbose flag to print output while updates are running rather than on error.
Adds test_prefix to specify a prefix of tests you want to run
Flushes stderr more reliably for builders and debugging purposes
Dies when stateful_updates fail

Change-Id: I9e4289deec3688c52b92763eee9334a8868f948e

BUG=chromium-os:6910
TEST=Ran with test_prefix and verbose on and off

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=53ac954

[sosa, 2010-12-03 22:33:28]

http://codereview.chromium.org/5573007/diff/2001/bin/cros_au_test_harness.py
File bin/cros_au_test_harness.py (right):

http://codereview.chromium.org/5573007/diff/2001/bin/cros_au_test_harness.py#...
bin/cros_au_test_harness.py:429: parser.add_option('--no_graphics',
action='store_true',
I just sorted this by long name

http://codereview.chromium.org/5573007/diff/2001/bin/cros_au_test_harness.py#...
bin/cros_au_test_harness.py:477: test_loader.testMethodPrefix =
options.test_prefix
Cleaned this up to make it easier to read and put common stuff together

[dgarrett, 2010-12-06 18:35:23]

http://codereview.chromium.org/5573007/diff/2001/bin/cros_au_test_harness.py
File bin/cros_au_test_harness.py (right):

http://codereview.chromium.org/5573007/diff/2001/bin/cros_au_test_harness.py#...
bin/cros_au_test_harness.py:454: if options.verbose: run_command = RunCommand
This isn't quite what I thought you meant when we discussed this. RunCommand and
RunCommandCaptureOutput don't have the same signatures, especially the return
value(s) are different.

RunCommand returns either the error code, or the output of the command, but not
both.

RunCommandCaptureOutput captures everything and returns all of it as a tuple.

The run_command calls above will fail with --verbose when trying to unpack the
results.

[sosa, 2010-12-07 01:25:00]

PTAL.  I tested this change :)

[dgarrett, 2010-12-07 01:41:10]

If passed, but only because you found a bug I introduced moving code around.

In _AttemptUpdateWithPayloadExpectedFailure, there is a block of code intended
to match the exception handler that shouldn't be.

The UpdateImage and UpdateUsingPayloads need to throw exceptions (fully
populated with stdout) for the two test cases testCorruptedUpdate, and
testInterruptedUpdate to work.

My broken indention in _AttemptUpdateWithPayloadExpectedFailure allowed the
tests to pass when they shouldn't have.

[sosa, 2010-12-07 22:33:05]

PTAL re-tested.

Changed to generate UpdateExceptions when RunCommand fails under verbose.

Also fixed indention bug

[dgarrett, 2010-12-08 19:36:05]

It still feels weird to me to have some tests fail if you set --verbose, but I
understand why.

Otherwise LGTM, and thanks for fixing my indention problem!

On 2010/12/07 22:33:05, sosa wrote:
> PTAL re-tested.
> 
> Changed to generate UpdateExceptions when RunCommand fails under verbose.
> 
> Also fixed indention bug