Taskcluster frequently failing trying to fetch Firefox profile

[gsnedders]

See the master branch https://github.com/web-platform-tests/wpt/runs/42387737071 or the tools/ integration tests https://github.com/web-platform-tests/wpt/pull/52455/checks?check_run_id=42357970828, both of which failed with:

Traceback (most recent call last):
  File "/home/test/web-platform-tests/./wpt", line 10, in <module>
    wpt.main()
  File "/home/test/web-platform-tests/tools/wpt/wpt.py", line 233, in main
    rv = script(*args, **kwargs)
  File "/home/test/web-platform-tests/tools/wpt/run.py", line 1007, in run
    setup_cls, wptrunner_kwargs = setup_wptrunner(venv, **kwargs)
  File "/home/test/web-platform-tests/tools/wpt/run.py", line 979, in setup_wptrunner
    setup_cls.setup(kwargs)
  File "/home/test/web-platform-tests/tools/wpt/run.py", line 246, in setup
    self.setup_kwargs(kwargs)
  File "/home/test/web-platform-tests/tools/wpt/run.py", line 318, in setup_kwargs
    prefs_root = self.browser.install_prefs(kwargs["binary"],
  File "/home/test/web-platform-tests/tools/wpt/browser.py", line 528, in install_prefs
    self.get_profile_github(version, channel, dest, rev)
  File "/home/test/web-platform-tests/tools/wpt/browser.py", line 484, in get_profile_github
    file_data[rel_path] = get_file_github("mozilla-firefox/firefox",
  File "/home/test/web-platform-tests/tools/wpt/browser.py", line 87, in get_file_github
    data: bytes = get(f"https://raw.githubusercontent.com/{repo}/{ref}/{path}").content  # type: ignore
  File "/home/test/web-platform-tests/tools/wpt/utils.py", line 109, in get
    resp.raise_for_status()
  File "/home/test/web-platform-tests/_venv3/lib/python3.10/site-packages/requests/models.py", line 1024, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 429 Client Error: Too Many Requests for url: https://raw.githubusercontent.com/mozilla-firefox/firefox/main/testing/profiles/base/user.js

[gsnedders]

We don't seem to do any rate-limited when fetching files from GitHub, which is quite possibly the problem.

[jonathan-j-lee]

This is causing wpt-firefox-nightly-* to fail regularly:

• https://github.com/web-platform-tests/wpt/pull/52650/checks?check_run_id=42512158059
• https://github.com/web-platform-tests/wpt/pull/52648/checks?check_run_id=42508394709
• https://github.com/web-platform-tests/wpt/runs/42389395235

@jgraham Would it be possible to vendor copies of profiles.json and user.js for WPT CI? Just guessing from the contents, but it doesn't look like they will change very frequently.

[gsnedders]

(Regression from https://github.com/web-platform-tests/wpt/commit/1af206eb5e753093f218749dc92e7c9d35473e2d / https://github.com/web-platform-tests/wpt/pull/52391.)

[gsnedders]

https://github.com/web-platform-tests/wpt/pulls?q=is%3Apr++status%3Afailure+created%3A2025-05-13..2025-05-20 shows a lot of CI failures of the past week.

[jrobbins]

Ping. Any update on this? It is showing up on https://ecosystem-infra-rotation.appspot.com/.

[jgraham]

Probably the easiest fix here is to provide a token when trying to do the fetch; the regression is a result of Gecko moving to GitHub (which doesn't support the workflow here as well as Mercurial did), and GitHub simultaneously clamping down on the unauthenticated rate limit.

[gsnedders]

@jgraham said he couldn't get this working via git-clone without it taking far too long.

For me, at least, this works:

git clone --no-checkout --depth=1 --filter=tree:0 https://github.com/mozilla-firefox/firefox
git -C firefox sparse-checkout set --cone testing/profiles
git -C firefox checkout

Gives:

gsnedders@gsnedders-margot tmp.ANYOFvrdI4 % time ./foo.sh 
+ git clone --no-checkout --depth=1 --filter=tree:0 https://github.com/mozilla-firefox/firefox
Cloning into 'firefox'...
remote: Enumerating objects: 1, done.
remote: Counting objects: 100% (1/1), done.
remote: Total 1 (delta 0), reused 0 (delta 0), pack-reused 0 (from 0)
Receiving objects: 100% (1/1), done.
+ git -C firefox sparse-checkout set --cone testing/profiles
+ git -C firefox checkout
remote: Enumerating objects: 34878, done.
remote: Counting objects: 100% (34878/34878), done.
remote: Compressing objects: 100% (22450/22450), done.
remote: Total 34878 (delta 642), reused 27049 (delta 408), pack-reused 0 (from 0)
Receiving objects: 100% (34878/34878), 12.58 MiB | 4.84 MiB/s, done.
Resolving deltas: 100% (642/642), done.
remote: Enumerating objects: 97, done.
remote: Counting objects: 100% (97/97), done.
remote: Compressing objects: 100% (95/95), done.
remote: Total 97 (delta 7), reused 23 (delta 1), pack-reused 0 (from 0)
Receiving objects: 100% (97/97), 437.17 KiB | 3.67 MiB/s, done.
Resolving deltas: 100% (7/7), done.
Your branch is up to date with 'origin/main'.
./foo.sh  0.82s user 0.67s system 25% cpu 5.807 total

[jrobbins]

Ping. Any additional update on this? It is showing up on https://ecosystem-infra-rotation.appspot.com/ again.

[jrobbins]

Ping. This is showing up on https://ecosystem-infra-rotation.appspot.com/ as an urgent bug that is old.

[jonathan-j-lee]

I haven't seen any blocked PRs recently, so I think it's safe to backlog this.