[sql] Check for open db before releasing cache.

Superceeded by https://codereview.chromium.org/1382283003

[sql] Check for open db before releasing cache.

BUG=537636, 489784

[Scott Hess - ex-Googler, 2015-09-30 17:11:41]

The CQ bit was checked by shess@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-09-30 17:12:04]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1381563003/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1381563003/1

[Scott Hess - ex-Googler, 2015-09-30 17:13:54]

shess@chromium.org changed reviewers:
+ rmcilroy@chromium.org

[Scott Hess - ex-Googler, 2015-09-30 17:13:54]

I verified this by simulating a failure in login_database.cc (to match the newer
bug's crash), and also by writing a transaction unit test to verify that
unwinding the object after db close would work.  Unfortunately, the unit test
isn't feasible to checkin, since I had to modify statement.cc to prevent the
code from failing out preparing the rollback statement against a closed
database.

[Scott Hess - ex-Googler, 2015-09-30 17:25:04]

https://codereview.chromium.org/1381563003/diff/1/sql/connection.cc
File sql/connection.cc (right):

https://codereview.chromium.org/1381563003/diff/1/sql/connection.cc#newcode530
sql/connection.cc:530: 
I didn't put:
  if (!is_open())
    return;
here because it felt like unwarranted overkill.  Looking at the callers,
Execute* explicitly checks db status, commit is probably an error to call with
closed db, and close clears the statement's connection reference.

That feels like a series of terrible justifications, but this is probably not
the time to re-architect everything.

[commit-bot: I haz the power, 2015-09-30 18:15:44]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-09-30 18:15:45]

Dry run: This issue passed the CQ dry run.

[Scott Hess - ex-Googler, 2015-09-30 22:56:24]

BTW - my original change got reverted because of a thread-safety problem in a
unit test.  So I'll slipstream this when relanding AGAIN, and of course you'll
get to look at that one...

[rmcilroy, 2015-10-01 10:48:37]

On 2015/09/30 22:56:24, Scott Hess wrote:
> BTW - my original change got reverted because of a thread-safety problem in a
> unit test.  So I'll slipstream this when relanding AGAIN, and of course you'll
> get to look at that one...

OK, I'll take a look in the reland.

[Scott Hess - ex-Googler, 2016-03-02 18:49:33]

Description was changed from

==========
[sql] Check for open db before releasing cache.

BUG=537636,489784
==========

to

==========
Superceeded by https://codereview.chromium.org/1382283003

[sql] Check for open db before releasing cache.

BUG=537636,489784
==========