latest update of browser compatibility data breaks tests
Categories
(DevTools :: Inspector: Compatibility, defect)
Tracking
(firefox-esr102 fixed, firefox102 fixed, firefox103 fixed, firefox104 fixed)
People
(Reporter: aryx, Assigned: jdescottes)
Details
Attachments
(1 file)
See this failure - browser compatibility data tests fail after the latest data update which had to be rolled back because of this.
| Reporter | ||
Comment 1•3 years ago
|
||
Daisuke, could you investigate this or redirect the request, please?
Comment 2•3 years ago
|
||
Thank you very much for the report, Sebastian!
Julian, as we talked on slack, could you take a look at this, please?
| Assignee | ||
Comment 3•3 years ago
|
||
The data dump removed IE, which fails a lot of expectations in our test suite.
For instance in browser_compatibility_css-property_issue.js, we expect 2 issues for the selected element, but one of them is for IE and since it's not a listed browser anymore, the issue is not displayed anymore.
Others are more straightforward such as browser_compatibility_settings.js where the list of browsers simply no longer matches.
Nicolas: do we have any process to address this scenario, when a data update breaks our existing tests? Or should we rewrite our tests to use a mock data set?
In the meantime, I can provide a patch to fix our tests with this data dump, but they will have to land together.
| Assignee | ||
Comment 4•3 years ago
|
||
I should actually have a version that works both with central and with the data dump, taking the bug.
| Assignee | ||
Comment 5•3 years ago
|
||
The compatibility panel tests should now work both with the current browser data and with the update from D150961
We will still need to find a reliable way to avoid breakage when a data dump occurs.
Comment 7•3 years ago
|
||
(In reply to Julian Descottes [:jdescottes] from comment #3)
The data dump removed IE, which fails a lot of expectations in our test suite.
For instance in browser_compatibility_css-property_issue.js, we expect 2 issues for the selected element, but one of them is for IE and since it's not a listed browser anymore, the issue is not displayed anymore.
Others are more straightforward such as browser_compatibility_settings.js where the list of browsers simply no longer matches.Nicolas: do we have any process to address this scenario, when a data update breaks our existing tests? Or should we rewrite our tests to use a mock data set?
Not really. I had an issue at some point because a property we were using in a test got supported by a browser and I fixed the test by using an other property. I can see that you used a similar approach in your patch, so we should probably be good
| Assignee | ||
Comment 8•3 years ago
|
||
That feels a bit flaky :) Ideally we should either run our tests with mocked data (or a fixed version of the data, to be updated periodically as a release task), or whatever process triggers the data dumps should also run our tests and involve us in case fixes are needed.
Comment 9•3 years ago
|
||
| bugherder | ||
Comment 10•3 years ago
|
||
FWIW, the search team has a similar issue when occasionally they need to update search configs and some tests in sync. When that happens they coordinate updating the config in remote-settings and the test changes with relman so nobody is surprised.
Comment 11•3 years ago
|
||
Oops didn't mean to set all those flags.
Comment 12•3 years ago
|
||
| bugherder uplift | ||
Comment 13•3 years ago
|
||
| bugherder uplift | ||
Comment 14•3 years ago
|
||
| bugherder uplift | ||
Description
•