Remove all the unused automation IPCs.

Remove all the unused automation IPCs.

BUG=121574
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=137672

[jam, 2012-05-17 16:20:33]

I'll update the reference builds once the builders are finished.

[Elliot Glaysher, 2012-05-17 16:30:05]

lgtm

[cmp, 2012-05-17 18:04:31]

It seems unnecessary to break compatibility when the change is just removing
lines.  I think this CL can be changed so that all the unused IPC messages can
be removed and each removed line replaced with a placeholder line to retain
compatibility.  This way the reference builds won't need to be updated.

[jam, 2012-05-17 18:10:29]

I agree that normally no need to do this, when one or two IPCs aren't used
anymore. But given that we've put it off for nearly a year, almost half of
the IPCs in automation_messages are now unused which seems excessive to me.

On Thu, May 17, 2012 at 11:04 AM, <cmp@chromium.org> wrote:

> It seems unnecessary to break compatibility when the change is just
> removing
> lines.  I think this CL can be changed so that all the unused IPC messages
> can
> be removed and each removed line replaced with a placeholder line to retain
> compatibility.  This way the reference builds won't need to be updated.
>
>
http://codereview.chromium.**org/10383232/<http://codereview.chromium.org/103...
>

[cmp, 2012-05-17 18:25:34]

What is the cost of putting off removing unused IPC messages in the reference
build?  AFAIK there is no cost, so if there is, then that could change my
opinion of what we do here.

The cost to updating the reference build is ensuring that the reference builds
are correctly updated, that tests on the trunk and the branch continue to work
after the update, and that perf expectations are all updated correctly and
working once more.  Selecting a good time to update the reference build is also
important, since now the performance of the reference build will change in our
perf tests.  Some tests may get more or less flaky, for example.  Some tests may
start to show the reference build's perf has greater variance in its results.