Sisyphus - Crash Automation - only update history bug_lists

RESOLVED FIXED

Status

Testing
Sisyphus
RESOLVED FIXED
8 years ago
8 years ago

People

(Reporter: bc, Assigned: bc)

Tracking

Trunk
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

Attachments

(3 attachments, 3 obsolete attachments)

(Assignee)

Description

8 years ago
Created attachment 437056 [details] [diff] [review]
patch v1

Currently each crash, assertion, valgrind result document in the crashtest or unittest database contains a bug_list field containing the open and closed bugs which match the result. This information is also maintained in the history database for each type.

Currently, the program attempts to keep the two in sync by updating the crash|assertion|valgrind result document in the test database and then updating the corresponding history document. This does not scale and is a continuing performance issue.

This patch changes the process to work from the history database and maintain the bug_list there. It temporarily retains the updates of the crash|assertion|valgrind results in the test database since the UI currently requires it. When the dependency of the UI on the bug_list in the crash|assertion|valgrind results is removed, this code can be eliminated.

Note this assumes the UI is completely responsible for maintaining the triage information (bug, comment) as it is impossible as far as I know for the code to make good decisions about triage.
Attachment #437056 - Flags: review?(ctalbert)
Attachment #437056 - Flags: feedback?(harthur)
(Assignee)

Comment 1

8 years ago
Created attachment 437325 [details] [diff] [review]
patch v2

make sure to do a full search if the bug_list was originally null.
Attachment #437056 - Attachment is obsolete: true
Attachment #437325 - Flags: review?(ctalbert)
Attachment #437325 - Flags: feedback?(harthur)
Attachment #437056 - Flags: review?(ctalbert)
Attachment #437056 - Flags: feedback?(harthur)
(Assignee)

Comment 2

8 years ago
Created attachment 438165 [details] [diff] [review]
patch v3
Attachment #437325 - Attachment is obsolete: true
Attachment #438165 - Flags: review?(ctalbert)
Attachment #437325 - Flags: review?(ctalbert)
Attachment #437325 - Flags: feedback?(harthur)
(Assignee)

Comment 3

8 years ago
Created attachment 438204 [details] [diff] [review]
patch v4

I stopped everyone but rebooting them and I guess I left a history document without a bug_list. This should safeguard that circumstance.
Attachment #438204 - Flags: review?(ctalbert)
(Assignee)

Updated

8 years ago
Attachment #438165 - Attachment is obsolete: true
Attachment #438165 - Flags: review?(ctalbert)
(Assignee)

Comment 4

8 years ago
Created attachment 438502 [details] [diff] [review]
follow on patch 1

bug_list isn't updated in the testdb's anymore and so update_bug_list_* don't need to return it.

Eliminate bug_list in the testdb's assertion|crash|valgrind documents.

Fix hold over search in update_bug_histories to point to the historydb and not the testdb.

Don't process HEAP valgrind messages.

Don't record special 'head' urls for crash on loads.

Don't try to get related bugs for signatures|stacks with only 0x0.
Attachment #438502 - Flags: review?(ctalbert)
(Assignee)

Comment 5

8 years ago
Created attachment 438504 [details] [diff] [review]
follow on patch 2

Since the unittest and crashtest workers can collide when updating assertions, etc in the history, only allow one to update histories at a time.
Attachment #438504 - Flags: review?(ctalbert)

Updated

8 years ago
Attachment #438204 - Flags: review?(ctalbert) → review+

Updated

8 years ago
Attachment #438502 - Flags: review?(ctalbert) → review+

Updated

8 years ago
Attachment #438504 - Flags: review?(ctalbert) → review+
(Assignee)

Comment 7

8 years ago
http://hg.mozilla.org/qa/sisyphus/rev/a9bdea79da9d
Status: ASSIGNED → RESOLVED
Last Resolved: 8 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.