Closed Bug 1528374 Opened 5 years ago Closed 5 years ago

Add macos pgo builds to taskcluster

Categories

(Firefox Build System :: General, enhancement)

enhancement
Not set
normal

Tracking

(firefox67 wontfix, firefox73 fixed)

RESOLVED FIXED
mozilla73
Tracking Status
firefox67 --- wontfix
firefox73 --- fixed

People

(Reporter: chmanchester, Assigned: chmanchester)

References

Details

Attachments

(5 files, 1 obsolete file)

This bug is about landing the initial implementation of macOS pgo builds in Taskcluster (maybe switched off by default to preserve resources). Anything that looks like shipping these will wait for bug 1515415.

These patches have been reviewed. Provided it will be available shortly I'll hold off landing this until bug 1530732 is ready so we don't burn time populating vcs caches in the general testing pool of macs.

Depends on: 1530732
Pushed by cmanchester@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/91c5e49bc1cb
Set artifact name based on platform in pgo profile run transform. r=firefox-build-system-reviewers,mshal
https://hg.mozilla.org/integration/autoland/rev/bf12f5e8abad
Convert binary argument to absolute path in profileserver.py r=firefox-build-system-reviewers,mshal
Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla67

This was just the first two patches in the stack so mshal could work on top of them, I should have marked this leave-open.

Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Depends on: 1499051

Just curious, are you planning to enable Rust PGO at the same time as C++? Or do you envision that being a separate bug?

Attachment #9045068 - Attachment is obsolete: true

(In reply to :dmajor from comment #11)

Just curious, are you planning to enable Rust PGO at the same time as C++? Or do you envision that being a separate bug?

I have it flipped on in the mozconfig, any reason not to do this here?

No objection from my side!

Pushed by cmanchester@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/a412a319534c
Find python3 correctly for 1014-pgo workers. r=tomprince
https://hg.mozilla.org/integration/autoland/rev/32f3b1b3fe3b
Add a simple script to wrap mozinstall. r=firefox-build-system-reviewers,mshal
https://hg.mozilla.org/integration/autoland/rev/5a6fa3b5123b
Add macOS pgo builds to the taskgraph. r=firefox-build-system-reviewers,mshal,tomprince
Backout by nerli@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/6a223e52eb0d
Backed out 3 changesets per tomprice's request for breaking macOS signing jobs CLOSED TREE

These ended up failing anyway, due to an issue with robustcheckout: https://treeherder.mozilla.org/#/jobs?repo=autoland&selectedJob=281011952

Connor, does that mean our mercurial is too new and we're getting python 3 despite an incompatible extension? Do we just need to update as in https://bugzilla.mozilla.org/show_bug.cgi?id=1561956#c35 ?

Flags: needinfo?(sheehan)

Sounds like mac workers have had mercurial updated to use python 3, but not robustcheckout.

(In reply to Mike Hommey [:glandium] (high latency) from comment #17)

Sounds like mac workers have had mercurial updated to use python 3, but not robustcheckout.

Yes, that's correct. Please grab the latest robustcheckout from version-control-tools and this problem will go away. I removed the urllib2 import in changeset c466f1409951 as part of the Python 3 compat upgrade.

Flags: needinfo?(sheehan)

Hi Tom, I'm confused by the reason given for backout. These are added as a new build job initially, where would that interfere with signing?

Flags: needinfo?(mozilla)

It was due to confusion on my part. Without looking at the patch, I had assumed that it was changing the shippable builds, and so would impact nightlies (I had paged most of the details of the in-tree part of this work out), and it was late, so I asked for it to be backed out to prevent issues with the next nightly. Given that I was incorrect about impact shippabe builds, there should not be an issue.

Flags: needinfo?(mozilla)

(In reply to Tom Prince [:tomprince] from comment #20)

It was due to confusion on my part. Without looking at the patch, I had assumed that it was changing the shippable builds, and so would impact nightlies (I had paged most of the details of the in-tree part of this work out), and it was late, so I asked for it to be backed out to prevent issues with the next nightly. Given that I was incorrect about impact shippabe builds, there should not be an issue.

Ok, great. I will attempt to re-land...

Pushed by cmanchester@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/0d0186ecd70e
Find python3 correctly for 1014-pgo workers. r=tomprince
https://hg.mozilla.org/integration/autoland/rev/967d0072cd2f
Add a simple script to wrap mozinstall. r=firefox-build-system-reviewers,mshal
https://hg.mozilla.org/integration/autoland/rev/3c2a1cf616b4
Add macOS pgo builds to the taskgraph. r=firefox-build-system-reviewers,mshal,tomprince

I had blindly reduced the timeout in a recent update. I will attempt re-landing with the old timeout...

Flags: needinfo?(cmanchester)
Pushed by cmanchester@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/d7c5e3a298ee
Find python3 correctly for 1014-pgo workers. r=tomprince
https://hg.mozilla.org/integration/autoland/rev/e71334caff30
Add a simple script to wrap mozinstall. r=firefox-build-system-reviewers,mshal
https://hg.mozilla.org/integration/autoland/rev/c99975fe700c
Add macOS pgo builds to the taskgraph. r=firefox-build-system-reviewers,mshal,tomprince
Status: REOPENED → RESOLVED
Closed: 5 years ago5 years ago
Resolution: --- → FIXED
Target Milestone: mozilla67 → mozilla73
Regressions: 1605226
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: