retry does not work for sub WF if failed manually

[Winson Chan]

In the example below, I'm expecting multiple executions of test-rollback.work and then error for test-rollback.main. Instead test-rollback.main stuck in RUNNING state and there's only 1 execution of test-rollback.work.

~$ cat ~/tmp/test-rollback.yaml
version: '2.0'
name: test-rollback

workflows:

    main:
        type: direct
        tasks:
            do:
                workflow: work
                retry:
                    count: 10
                    delay: 1

    work:
        type: direct
        tasks:
            do:
                action: std.fail
                on-error:
                    - undo
            undo:
                action: std.echo output="rolling back..."
                on-complete:
                    - fail

~$ mistral workbook-create ~/tmp/test-rollback.yaml
Starting new HTTP connection (1): localhost
+------------+----------------------------+
| Field | Value |
+------------+----------------------------+
| Name | test-rollback |
| Tags | <none> |
| Created at | 2015-04-27 23:06:39.866864 |
| Updated at | None |
+------------+----------------------------+

~$ mistral execution-create test-rollback.main
Starting new HTTP connection (1): localhost
+------------+--------------------------------------+
| Field | Value |
+------------+--------------------------------------+
| ID | d67824fc-52bb-49e2-b736-067fc4ca26a6 |
| Workflow | test-rollback.main |
| State | RUNNING |
| Created at | 2015-04-27 23:06:49.022015 |
| Updated at | 2015-04-27 23:06:49.216968 |
+------------+--------------------------------------+

~$ mistral execution-list
Starting new HTTP connection (1): localhost
+--------------------------------------+------------------------------------+---------+----------------------------+----------------------------+
| ID | Workflow | State | Created at | Updated at |
+--------------------------------------+------------------------------------+---------+----------------------------+----------------------------+
| d67824fc-52bb-49e2-b736-067fc4ca26a6 | test-rollback.main | RUNNING | 2015-04-27 23:06:49.022015 | 2015-04-27 23:06:49.216968 |
| 75958b6b-444d-4851-92b9-14c3c7741354 | test-rollback.work | ERROR | 2015-04-27 23:06:49.675953 | 2015-04-27 23:06:52.452506 |
+--------------------------------------+------------------------------------+---------+----------------------------+----------------------------+

~$ mistral task-list 75958b6b-444d-4851-92b9-14c3c7741354
Starting new HTTP connection (1): localhost
+--------------------------------------+------+--------------------+--------------------------------------+---------+
| ID | Name | Workflow name | Execution ID | State |
+--------------------------------------+------+--------------------+--------------------------------------+---------+
| adc2eee2-a499-426f-9ac3-695a30256f4f | do | test-rollback.work | 75958b6b-444d-4851-92b9-14c3c7741354 | ERROR |
| dd09f069-1789-4c67-8100-257eab9ade9b | undo | test-rollback.work | 75958b6b-444d-4851-92b9-14c3c7741354 | SUCCESS |
+--------------------------------------+------+--------------------+--------------------------------------+---------+

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/186053

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/186053
Committed: https://git.openstack.org/cgit/stackforge/mistral/commit/?id=766698bd438b070cc32fa2a15cf033eddefee3c4
Submitter: Jenkins
Branch: master

commit 766698bd438b070cc32fa2a15cf033eddefee3c4
Author: Nikolay Mahotkin <email address hidden>
Date: Tue May 26 16:01:48 2015 +0300

    Fixing sending the result of subworkflow

     * This is why retry didn't work for subworkflow
     * Engine commands such as fail or succeed doesn't send the
       result of current workflow to parent workflow if needed.

    Closes-Bug: #1449298

    Change-Id: Ia0d87160a51f53a686ca0451a2000298cec5a929