Closed
Bug 1465181
Opened 5 years ago
Closed 5 years ago
Support native-engine in run-task
Categories
(Firefox Build System :: Task Configuration, task)
Tracking
(firefox62 fixed)
RESOLVED
FIXED
mozilla62
Tracking | Status | |
---|---|---|
firefox62 | --- | fixed |
People
(Reporter: ahal, Assigned: ahal)
References
Details
Attachments
(4 files)
Native engine is nominally already supported: https://searchfox.org/mozilla-central/source/taskcluster/taskgraph/transforms/job/run_task.py#111 But when I created a task that tried to use it, I ran into a bunch of issues. Mostly revolving around the fact that the task isn't being run as root. Fixing this will allow us to run the new js-bench tasks on hardware as opposed to AWS.
Assignee | ||
Updated•5 years ago
|
Assignee: nobody → ahal
Status: NEW → ASSIGNED
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment 5•5 years ago
|
||
mozreview-review |
Comment on attachment 8981625 [details] Bug 1465181 - [taskgraph] Stop hardcoding the workdir to /builds/worker in 'job' tasks, https://reviewboard.mozilla.org/r/247746/#review253772 The bulk of the changes seem fine. I'm just a bit unnerved by the dynamic schema modification. ::: taskcluster/taskgraph/transforms/job/__init__.py:183 (Diff revision 1) > > The decorated function should have the signature `using_foo(config, job, > taskdesc) and should modify the task description in-place. The skeleton of > the task description is already set up, but without a payload.""" > + if schema: > + schema = schema.extend({Optional('workdir'): basestring}) Modifying the schema here doesn't seem correct. Shouldn't the schema be defined in a constant somewhere? If we do dynamic schema modification elsewhere, I'm inclined to turn a blind eye. But this feels not right.
Attachment #8981625 -
Flags: review?(gps) → review-
Comment 6•5 years ago
|
||
mozreview-review |
Comment on attachment 8981626 [details] Bug 1465181 - [run-task] Remove requirement to run as root on POSIX systems, https://reviewboard.mozilla.org/r/247748/#review253774 Yes, the check is superficial.
Attachment #8981626 -
Flags: review?(gps) → review+
Assignee | ||
Comment 7•5 years ago
|
||
mozreview-review-reply |
Comment on attachment 8981625 [details] Bug 1465181 - [taskgraph] Stop hardcoding the workdir to /builds/worker in 'job' tasks, https://reviewboard.mozilla.org/r/247746/#review253772 > Modifying the schema here doesn't seem correct. Shouldn't the schema be defined in a constant somewhere? > > If we do dynamic schema modification elsewhere, I'm inclined to turn a blind eye. But this feels not right. Good point. The fact that schemas are duplicated all over the place is one of the things that annoy me about taskgraph, but this bug is the wrong hill to die on. It should be solved properly in the general case rather than having one off schema modification.
Comment 8•5 years ago
|
||
mozreview-review |
Comment on attachment 8981627 [details] Bug 1465181 - Use releng-hardware/gecko-t-linux-talos workerType for js-bench tasks, https://reviewboard.mozilla.org/r/247750/#review253970 thanks
Attachment #8981627 -
Flags: review?(jmaher) → review+
Assignee | ||
Comment 9•5 years ago
|
||
The last commit in this series is going to conflict with bug 1465336 (which just landed on inbound).
Depends on: 1465336
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment 14•5 years ago
|
||
mozreview-review |
Comment on attachment 8981625 [details] Bug 1465181 - [taskgraph] Stop hardcoding the workdir to /builds/worker in 'job' tasks, https://reviewboard.mozilla.org/r/247746/#review254134 FWIW, we still want to strive for using `/builds/worker` in as many tasks as possible, especially on the build side. That's because things like sccache cache according to the absolute path. Just something to keep in mind. Talos should be immune from these concerns.
Attachment #8981625 -
Flags: review?(gps) → review+
Assignee | ||
Comment 15•5 years ago
|
||
Good to know! The only reason I did this was because the jsshell task I was setting up didn't have permission to write to /builds/worker, and it doesn't do any compiling. Fwiw, I don't plan on using this feature again (unless a similar situation happens in the future).
Comment 16•5 years ago
|
||
mozreview-review |
Comment on attachment 8981624 [details] Bug 1465181 - [taskgraph] Support use-artifacts with native-engine in run-task, https://reviewboard.mozilla.org/r/247744/#review254428 sorry for the review delay
Attachment #8981624 -
Flags: review?(dustin) → review+
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment hidden (mozreview-request) |
Comment 21•5 years ago
|
||
Pushed by ahalberstadt@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/f84d6a6ce047 [taskgraph] Support use-artifacts with native-engine in run-task, r=dustin https://hg.mozilla.org/integration/autoland/rev/2c68830a03a0 [taskgraph] Stop hardcoding the workdir to /builds/worker in 'job' tasks, r=gps https://hg.mozilla.org/integration/autoland/rev/f75a5959d692 [run-task] Remove requirement to run as root on POSIX systems, r=gps https://hg.mozilla.org/integration/autoland/rev/71ab37336d27 Use releng-hardware/gecko-t-linux-talos workerType for js-bench tasks, r=jmaher
Comment 22•5 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/f84d6a6ce047 https://hg.mozilla.org/mozilla-central/rev/2c68830a03a0 https://hg.mozilla.org/mozilla-central/rev/f75a5959d692 https://hg.mozilla.org/mozilla-central/rev/71ab37336d27
Status: ASSIGNED → RESOLVED
Closed: 5 years ago
status-firefox62:
--- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla62
Updated•4 years ago
|
Version: Version 3 → 3 Branch
You need to log in
before you can comment on or make changes to this bug.
Description
•