Add ThreadPool.abort() to stop processing early.

Add ThreadPool.abort() to stop processing early.

Make Progress more responsive in ThreadPool.join().
Change update_item() to allow decrement.

R=vadimsh@chromium.org
BUG=

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=226540

[Marc-Antoine Ruel (Google), 2013-10-02 14:08:31]

I needed .abort() for my load test. And I use the decrement size=-1 in the load
test too.

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py
File utils/threading_utils.py (right):

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py#n...
utils/threading_utils.py:389: def update_item(self, name, index=0, size=0,
raw=False):
Eventually I want it to support an arbitrary number of column. At least I have a
use case for 3 numbers. That's out of scope for this CL.

[Vadim Sh., 2013-10-02 18:19:48]

https://codereview.chromium.org/25478010/diff/3001/googletest/run_test_cases.py
File googletest/run_test_cases.py (right):

https://codereview.chromium.org/25478010/diff/3001/googletest/run_test_cases....
googletest/run_test_cases.py:867: self.progress.update_item(line, index=1,
size=need_to_retry)
int(need_to_retry)?

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py
File utils/threading_utils.py (right):

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py#n...
utils/threading_utils.py:509: self.all_tasks_done.wait(0.1)
I'm unhappy about polling. I think print_update can be rewritten to use blocking
get from self.queued_lines. Then updates will be printed instantly and
performance will also be better.

For instance in currently implementation all log handlers will be flushed each
100ms, it's not great.

But it's probably for some future CL.

[Marc-Antoine Ruel (Google), 2013-10-02 19:41:26]

https://codereview.chromium.org/25478010/diff/3001/googletest/run_test_cases.py
File googletest/run_test_cases.py (right):

https://codereview.chromium.org/25478010/diff/3001/googletest/run_test_cases....
googletest/run_test_cases.py:867: self.progress.update_item(line, index=1,
size=need_to_retry)
On 2013/10/02 18:19:48, Vadim Sh. wrote:
> int(need_to_retry)?

Done.

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py
File utils/threading_utils.py (right):

https://codereview.chromium.org/25478010/diff/3001/utils/threading_utils.py#n...
utils/threading_utils.py:509: self.all_tasks_done.wait(0.1)
On 2013/10/02 18:19:48, Vadim Sh. wrote:
> I'm unhappy about polling. I think print_update can be rewritten to use
blocking
> get from self.queued_lines. Then updates will be printed instantly and
> performance will also be better.
> 
> For instance in currently implementation all log handlers will be flushed each
> 100ms, it's not great.
> 
> But it's probably for some future CL.

I agree it's far from awesome but I prefer the Progress to stay "interactive".
So I added a TODO.

[Vadim Sh., 2013-10-02 19:47:52]

ok.. lgtm

[commit-bot: I haz the power, 2013-10-02 20:00:29]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/maruel@chromium.org/25478010/5001

[commit-bot: I haz the power, 2013-10-02 20:02:10]

Change committed as 226540