GitLab

This change introduces support for interrupts.

Interrupts can be run as part of wait_for_dom with
allow_interrupts=true.

Interrupts can also run manually, by clicking on a chip, but only if
they have a name.

Interrupts can run automatically if they have autostart=true. Autostart
interrupts are handled differently from normal autostart, as they can
happen at any time. Running an autostart interrupt does not prevent
runnig a normal interrupt.

I tried to keep changes to the main, non-interrupt, path and
refactorings to a minimum in this change.

The result puts all the complexity into the script executor. I'd like to
get it to work, then find a good code split in a follow-up change. The
code has resisted all my attempts at reorganization so far...

Change to the main, non-interrupt path:

 - The responsibility for updating the script status is moved to
 ScriptExecutor, even though the script status map stays in
 ScriptTracker.This allows tracking script state changes caused by
 interrupts running.

 - The WaitForDom now uses the same wait code as
 ActionDelegate::WaitForElement instead of rolling up its own

Limitations:

 - The payload from interrupts is never sent back to the server. This
 doesn't allow interrupts to change variables

 - If an interrupt fails, the failure of the main script is not
 reported to the server; from the server's point of view, the main
 script just never finishes.

 - There's no "prompt" action yet, so interrupts cannot ask questions.

Bug: 806868
Change-Id: I12165965783ba479688c4c5b44ab72477ad4ae2b
Reviewed-on: https://chromium-review.googlesource.com/c/1340309
Commit-Queue: Stephane Zermatten <szermatt@chromium.org>
Reviewed-by: Mathias Carlen <mcarlen@chromium.org>
Cr-Commit-Position: refs/heads/master@{#610382}