Always use the topmost delayed task as the basis for our WM_TIMER....

Always use the topmost delayed task as the basis for our WM_TIMER.

R=jar
BUG=2559
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=2509

[darin (slow to review), 2008-09-23 15:45:57]

Note: I ended up writing a test for this that could be potentially flaky if a 10
seconds suddenly evaporated during test execution.  I believe that window of
opportunity to be significantly low enough to not worry about it.

[jar (doing other things), 2008-09-23 18:21:15]

Patch for timer looks LGTM

Test is LGTM, but perhaps (per discussion) we could use a larger time, and
measure the run time to verify it is short duration before the 10ms task runs.

Nice job with subpump test!!!

(p.s., Duration EXPECT() code should be in both sub-pump and regular pump tests)

http://codereview.chromium.org/4041/diff/1/2
File base/message_loop_unittest.cc (right):

http://codereview.chromium.org/4041/diff/1/2#newcode305
Line 305: FROM_HERE, new RecordRunTimeTask(&run_time1, &num_tasks), 10000);
Per discussion: I'd suggest using a longer time value here, and checking that we
ran in well under some nice number of seconds below.  You may want to comment
that IF the test harness times us out, that is a clear failure of this test.

[jar (doing other things), 2008-09-23 18:49:48]

LGTM