Base test class for chrome tests

Base test class for chrome tests
Simplifies wrapping of existing chrome tests for autotest environment.

Change-Id: Ifd7fb3aae48fc080ea3f67e13aaff3a7f9d5ffda

BUG=none
TEST=none

[asharif1, 2010-09-15 20:37:05]

On 2010/09/15 19:00:09, zel wrote:
> 

This is *almost* what I need for my test.

I would say please add an option to record the stdout or stderr of the command. 

In page_cycler, I read/parse/interpret it.

Also (perhaps page_cycler is too specialized of a test),  perhaps it would be
nice to be able to place other stuff in the CR_SOURCE_ROOT dir, like tarballs,
etc. Page_cycler needs to extract a tarball to this dir (or at least put some
symlinks here). 

Lastly, about generating a random dir, the best thing would to be run mktemp -d
to get a temp dir instead of generating a random number. The dir *can* conflict
with your approach and mktemp is safer.

Otherwise LGTM.

[asharif1, 2010-09-15 20:48:42]

On 2010/09/15 20:37:05, asharif1 wrote:
> On 2010/09/15 19:00:09, zel wrote:
> > 
> 
> This is *almost* what I need for my test.
> 
> I would say please add an option to record the stdout or stderr of the
command. 
> 
> In page_cycler, I read/parse/interpret it.
> 
> Also (perhaps page_cycler is too specialized of a test),  perhaps it would be
> nice to be able to place other stuff in the CR_SOURCE_ROOT dir, like tarballs,
> etc. Page_cycler needs to extract a tarball to this dir (or at least put some
> symlinks here). 
> 
> Lastly, about generating a random dir, the best thing would to be run mktemp
-d
> to get a temp dir instead of generating a random number. The dir *can*
conflict
> with your approach and mktemp is safer.
> 
> Otherwise LGTM.

To be completely general, maybe you can add an option of passing in a function
pointer to this command so it can run an arbitrary function with the output of
mktemp.

The function can do anything to the root, for example, extract a tarball there
or touch a file there, etc. 

An alternative would be to have two functions: the 1st to setup the source and
return the folder where it setup the source, and the 2nd to actually run the
test and capture the output/error stream of it.

A test can potentially write some stuff to it's directory and it would be nice
to read it back by using the saved dir from function 1.

At the end we can call a function that cleans up the temp dir.

[zel, 2010-09-15 22:05:57]

I have improved temp dir handling there and would like to get this checked so
the test actually start working.

I am intending to add additional features there in the next set of CLs and will
be happy to look into page_cycler related scenario as well.

[asharif1, 2010-09-15 22:18:08]

On 2010/09/15 22:05:57, zel wrote:
> I have improved temp dir handling there and would like to get this checked so
> the test actually start working.
> 
> I am intending to add additional features there in the next set of CLs and
will
> be happy to look into page_cycler related scenario as well.

k, I have a few other features to request.

I am not sure whether this is your domain (I think it is), but I would like to
make the chrome_test dep faster. Right now it bundles a lot of stuff I don't
need. Perhaps it can be smarter in terms of bundling if the user only wants a
single binary (for example, ideally I would like to only bundle the
page_cycler_tests binary which is around 50 megs, and not the other hundreds of
megs of other tests). This makes writing autotests a big pain because of the
long compile-edit-run cycle.

LGTM for now. Keep these features in mind when submitting future CLs.

I will not be using this code for now (for page_cycler) until the improvements
come in. I'll add you as a reviewer for that CL.

[kmixter1, 2010-09-15 22:29:16]

LGTM with some comments...

http://codereview.chromium.org/3403006/diff/7001/8001
File client/common_lib/chrome_test_base.py (right):

http://codereview.chromium.org/3403006/diff/7001/8001#newcode1
client/common_lib/chrome_test_base.py:1: # Copyright (c) 2010 The Chromium OS
Authors. All rights reserved.
I think the convention is to call these site_*_test.py.  See
site_{ui,crash}_test.py in bin directory.

http://codereview.chromium.org/3403006/diff/7001/8001#newcode9
client/common_lib/chrome_test_base.py:9: class chrome_test_base(test.test):
Classes should be camel case

http://codereview.chromium.org/3403006/diff/7001/8001#newcode41
client/common_lib/chrome_test_base.py:41: shutil.rmtree(home_dir,
ignore_errors=True)
probably should be in cleanup() so any intervening failures cause this directory
to disappear.

[zel, 2010-09-15 22:57:55]

http://codereview.chromium.org/3403006/diff/7001/8001
File client/common_lib/chrome_test_base.py (right):

http://codereview.chromium.org/3403006/diff/7001/8001#newcode1
client/common_lib/chrome_test_base.py:1: # Copyright (c) 2010 The Chromium OS
Authors. All rights reserved.
On 2010/09/15 22:29:16, kmixter1 wrote:
> I think the convention is to call these site_*_test.py.  See
> site_{ui,crash}_test.py in bin directory.

Done.

http://codereview.chromium.org/3403006/diff/7001/8001#newcode9
client/common_lib/chrome_test_base.py:9: class chrome_test_base(test.test):
On 2010/09/15 22:29:16, kmixter1 wrote:
> Classes should be camel case

Done.