Add trial_dir to manage automatic trial directory creation and deletion.

Add trial_dir to manage automatic trial directory creation and deletion.

The tests now base the directory in mkdtemp() which is much safer than using the
source's file directory as before.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=77278

[M-A Ruel, 2011-03-04 19:35:13]

This should help a lot.

Based on http://codereview.chromium.org/6632007.

[Evan Martin, 2011-03-04 19:46:11]

http://codereview.chromium.org/6625018/diff/1/tests/fake_repos.py
File tests/fake_repos.py (right):

http://codereview.chromium.org/6625018/diff/1/tests/fake_repos.py#newcode276
tests/fake_repos.py:276: super(FakeReposBase, self).set_up()
I feel like this mechanism, where you conditionally call the superclass, is
fragile.  Why not just use a has-a relationship for the FakeRepos class?  You
can call its set_up and tear_down directly...

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py
File tests/trial_dir.py (right):

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode6
tests/trial_dir.py:6: """Base unittest class that cleans off a trial
directory."""
Can you elaborate on this?  (English note: "cleans off" means more like cleaning
off dirt from a car than it means to remove a directory; maybe "cleans up after"
is better.)

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode16
tests/trial_dir.py:16: sys.path.insert(0, os.path.dirname(ROOT_DIR))
Isn't this implicit in how Python imports work?

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode43
tests/trial_dir.py:43: Note that it deletes all trial_dir() and not only
repos_dir.
repos_dir isn't meaningful in this file, I'm not sure what you meant here

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode87
tests/trial_dir.py:87: if '-l' in sys.argv:
This feels fragile, but I'll let you judge.  gyp uses environment variables to
control this kind of thing.

[M-A Ruel, 2011-03-04 20:00:49]

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py
File tests/trial_dir.py (right):

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode16
tests/trial_dir.py:16: sys.path.insert(0, os.path.dirname(ROOT_DIR))
On 2011/03/04 19:46:11, Evan Martin wrote:
> Isn't this implicit in how Python imports work?

No because I'm adding the parent's directory.

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode43
tests/trial_dir.py:43: Note that it deletes all trial_dir() and not only
repos_dir.
On 2011/03/04 19:46:11, Evan Martin wrote:
> repos_dir isn't meaningful in this file, I'm not sure what you meant here

Copy-paste fail.

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode87
tests/trial_dir.py:87: if '-l' in sys.argv:
On 2011/03/04 19:46:11, Evan Martin wrote:
> This feels fragile, but I'll let you judge.  gyp uses environment variables to
> control this kind of thing.

It's only used when testing manually to inspect the data when a test fails.

[Evan Martin, 2011-03-04 21:44:28]

sorry for slow review, didn't realize you'd uploaded fixes

LGTM, minor fix below

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py
File tests/trial_dir.py (right):

http://codereview.chromium.org/6625018/diff/1/tests/trial_dir.py#newcode6
tests/trial_dir.py:6: """Base unittest class that cleans off a trial
directory."""
This comment is now wrong, too.  :)

"""Helper class that manages a temporary directory for use in tests."""

[M-A Ruel, 2011-03-04 21:55:08]

Fixed the docstring.

[M-A Ruel, 2011-03-04 22:12:15]

On 2011/03/04 21:55:08, Marc-Antoine Ruel wrote:
> Fixed the docstring.

and fixed typo.

[commit-bot: I haz the power, 2011-03-08 13:40:31]

Change committed as 77278