Broken builds and merge-ups

[github-actions]

This is an automatically created issue used to remind the Silverstripe CMS Squad to look at broken builds and merge-ups every Tuesday.

It was created by the .github/workflows/rhino-issue.yml workflow in the silverstripe/.github repository.

Triage instructions

1. Put on the following label:

• type/other

2. Move this issue to the "Ready" column on our internal zenhub board

Broken builds:

• https://rhino.silverstripe.org/?t=builds

Merge-ups:

• https://rhino.silverstripe.org/?t=merge-ups

PRs

• MNT Add ci.yml so merge-up and patch tagging works as expected documentation-lint#7
• FIX Correctly set img tag attributes for WYSIWYG silverstripe-admin#1798
• MNT Test empty alt tags are included silverstripe-cms#2978
• FIX Don't cause errors if elemental isn't installed silverstripe-frameworktest#194
• FIX Don't require classes that don't exist silverstripe-frameworktest#195
• FIX Only include extension if elemental is installed silverstripe-frameworktest#196
• MNT Reuse existing classes for behat test silverstripe-elemental#1226
• ENH Don't use keyword "self" silverstripe-elemental-bannerblock#155
• ENH Don't use keywords self silverstripe-framework#11313
• ENH Don't use keyword self tractorcow-farm/silverstripe-fluent#867
• ENH Don't use keyword "self" silverstripe-assets#622
• DEP Conflict with incompatible version of elemental silverstripe-admin#1800
• MNT Fix unit test silverstripe-framework#11314
• ENH Don't use the keyword "self" silverstripe-framework#11315
• FIX Lock chrome and chromedriver to v126 until we know why 127 fails gha-ci#140
• ENH Don't use keyword "self" symbiote/silverstripe-gridfieldextensions#410

[GuySartorelli]

There's a weird problem affecting both asset admin CI and admin CI which I can't reproduce locally, where it can't find the dialog.

My initial thoughts are that it could be a timing problem, since other alerts do fire and are found where they're expected. Though it is weird for a timing problem to be so consistent.

Alternatively it could be a different version of chromium or chrome driver perhaps? Or some other version mismatch? Almost certainly infrastructure since it's consistent even in sink, installer, and graphql CI runs but I can't reproduce locally either with DDEV or with my old custom docker setup.

Edit: Looking at the user experience, and based on one of the failed scenarios (haven't checked the others yet), if "I dismiss the dialog" in CI is actually confirming the dialog, then that could explain why the second dialog doesn't appear.

[emteknetnz]

Assigning back to Guy to ensure things are working as they should

[GuySartorelli]

Gonna add relevant findings here as I go, both for my own tracking and for future reference.

• That asset-admin failure fails when just running that scenario on its own: https://github.com/creative-commoners/silverstripe-installer/actions/runs/10155247714. This rules out any bleed-over from other scenarios or features.

• It also fails when using admin 5.2, which means changes in that repo at least aren't to blame (since the ci failure doesn't happen in the 5.2 branch of asset-admin)

• It fails when using installer 5.2 with asset-admin 2.x-dev, which implies a code change in asset-admin specifically is to blame. Though that seems unlikely given a similar failure is affecting admin...

• Looks like the fact that 2.2.x-dev was passing is just 'cause that hadn't been run recently enough.

  • There are a few differences in dependencies so I'll check those first.
  • There's also differences in chromedriver and chromium versions which I really hope aren't the problem, but they probably are
  • admin 2.2 passing 2 weeks ago
    • Chrome version is: Google Chrome for Testing 126.0.6478.182
    • Chromedriver version is: ChromeDriver 126.0.6478.182 (5b5d8292ddf182f8b2096fa665b473b6317906d5-refs/branch-heads/6478@{#1776})
  • admin 2.2 failing more recently
    • Google Chrome for Testing 127.0.6533.72
    • ChromeDriver 127.0.6533.72 (9755e24ca85aa18ffa16c743f660a3d914902775-refs/branch-heads/6533@{#1760})

Turns out it was the chrome or chromedriver version causing problems. I can reproduce with local behat now.
Can't reproduce in-browser locally. Even downloaded the exact version of "chrome for testing" that's failing in CI and the failure isn't happening in-browser.

[GuySartorelli]

Opened php-webdriver/php-webdriver#1104 about the remaining behat failures in admin and asset-admin