Fix Chrome 72 extensions not working correctly (e.g. adblockers)
Chrome 72 has a confirmed bug currently that affects a small part of the overall population that uses the stable version of the browser.
Affected users may notice that extensions that work with connections, e.g. content blockers, work incorrectly on some sites.
The issue was reported on February 13 on the official Chromium bug listing site. According to the report, most requests from a Chrome 72 browser were not intercepted on some sites such as Gmail while blocking worked fine on others.
Extensions with confirmed issues are uBlock Origin and Mailtrack. Most content blockers are probably affected by the issue.
It turned out quickly that a variation of the Chrome configuration was responsible for the behavior. Google runs so-called Field Trials in Chrome to test new features and changes.
You can list all variations of a particular version of Chrome by loading chrome://version/.
The Field Trial responsible for the issue is called Network Service by Chromium engineers.The Network Service "makes network requests through a separate process" according to Google.
Google fixed the bug in Chrome 73 Beta but won't bring the fix to the current stable version of Chrome. In other words: Chrome 72 users who experience the issue need to find another way to resolve it.
Google suggests to upgrade to Chrome 73 Beta or disable the experiment. Considering that most users may not want to upgrade to a beta browser, here is how you disable the feature right now in Chrome 72:
- Load chrome://flags/#network-service in the Chrome address bar.
- Set the experiment to Disabled.
- Restart Chrome.
The issue should be fixed in Chrome Stable after the restart. Try connecting to Gmail or any other site that uses web workers to test if that is indeed the case.
Google decided not to stop the rollout of the experiment or roll it back.
At this point, if this is the only breakage in the experiment we're not rolling back. The reason is that as a relative percentage of users of Chrome, this is still small (e.g. less than 0.1%). When launching multi-year projects that impact a large part of the codebase, it's impossible to avoid any regressions. We have to balance making forward progress and avoiding other regressions creeping in with breaking some edge cases. The best way for extension authors to avoid this is to use dev/beta channels.
Google stating that extensions authors could avoid issues like the one experienced by testing against Dev and Beta versions is misleading in this case considering that the experiment might not even have been enabled in a particular version of Chrome.
While less than 0.1% of all Chrome users seems relatively low, it could still affect hundred of thousands of users.
Now You: Did you run in the issue? Should Google change its approach to tests?Advertisement