Getting a ? on mobile PageSpeed score

no-liver commented 3 years ago

Hi,

Our site is getting a ? mobile PageSpeed score. This issue arised a few days ago. Currently unable to pinpoint the issue from the site side, is there any possibility to assist?

Please view site: https://tinyurl.com/yyfx7kgl

Best

patrickhulce commented 3 years ago

Thanks for filing @rulloliver! I can reproduce this in PSI but no other environment. We'll have to look into it.

connorjclark commented 3 years ago

weird issue with the css protocol. attached JSON for the LHR/artifacts from LR for later analysis.

lrdata.txt

paulirish commented 3 years ago

connorjclark commented 3 years ago

perhaps related to #10876

I don't think so. I think the root cause is whatever caused SPEEDINDEX_OF_ZERO, aka no screenshots, for some reason ... Nothing painted = whole page was too busy = no CSS parsed yet ???

https://sentry.io/organizations/google-lighthouse/issues/1454681293/?project=174697&query=is%3Aunresolved+No+style+sheet&statsPeriod=14d

many (but not all) have a NO_TTI or NO_LCP error too..

But this suggests it's not a PSI issue.

I can't get this to happen anymore with the provided URL. @rulloliver have you since changed the page in a significant way? Perhaps made it use less JS?

brendankenny commented 3 years ago

perhaps related to #10876

I don't think so. I think the root cause is whatever caused SPEEDINDEX_OF_ZERO

I believe I suggested that comment in the last bug bash. It was just based off the CSSUsage/"no style sheet with given id found" error in the screenshot.

Nothing painted = whole page was too busy = no CSS parsed yet ???

We only add a style sheet to be queried when the protocol indicates one has been added (CSS.styleSheetAdded), not on network records. I would assume it would only generate the event/id after it has something that can be queried?

Lots of interesting weird things about these artifacts, though (here's a live version of the report from them)

It looks like the screenshot in https://github.com/GoogleChrome/lighthouse/issues/11385#issuecomment-687528372 doesn't match the attached artifacts, because those artifacts produce successful css-usage and TTI audits, even though you still get a SPEEDINDEX_OF_ZERO.
There are screenshots in the trace, they're just five screenshots (three after FCP) of all white. Hence the zero.
One possible bug for us here: it's questionable whether SPEEDINDEX_OF_ZERO should use the "Chrome didn't collect any screenshots" string, because it did collect screenshots.
That trace has an FCP (what was it?) and an LCP (which starts with the contents "By clicking \"Accept All Cookies\", you agree to the storing of cookies on your d…"), but where are they in the screenshots? Nothing appears to be painted.
several net::ERR_BLOCKED_BY_CLIENT.Inspector errors in there (#10362) and a document.write error via googleadservices and a js error on attempting to reference myLinkToDecorate

I'm not sure there's much for us to do here without a test case, though. It may just be there were enough errors in the site/PSI combo to not paint anything screenshottable but it was still enough to trigger the FCP/LCP thresholds.

no-liver commented 3 years ago

Issue resolved currently by itself but lasted for about a week. Haven't been able to replicate it since. Thank you for the feedback and help

connorjclark commented 2 years ago

One possible bug for us here: it's questionable whether SPEEDINDEX_OF_ZERO should use the "Chrome didn't collect any screenshots" string, because it did collect screenshots.

https://github.com/GoogleChrome/lighthouse/blob/d9e84e1d71e04ddae4f53d8e013683d6ca4c17ce/lighthouse-core/computed/speedline.js#L45

should we just check for undefined here, and allow for 0 to continue without error or throw a unique error for it?

GoogleChrome / lighthouse

Getting a ? on mobile PageSpeed score #11385