[BUG] [compression] Avoid enormous network payloads / Defer unused CSS doesn't consider compression

1,072 views
Skip to first unread message

Raul Peixoto

unread,
Feb 2, 2019, 12:53:49 AM2/2/19
to pagespeed-insights-discuss
I think there was a recent update, within the past week or so that started causing this AND it's affecting all scores on most wordpress sites, tremendously.

In a specific case, I have an 1.5 MB CSS file with brotli or gzip compression.

Screenshot_1.jpg

I can see that Chrome is downloading the correct br file:

Screenshot_2.jpg


Screenshot_3.jpg

And so does Chrome Lighthouse Version 71.0.3578.98:

Screenshot_4.jpg

it looks like this:

Screenshot_5.jpg

Screenshot_6.jpg

So it seems that suddenly, the PSI tool decided not to download compressed tools with either brotli or gzip.
This also affects https://web.dev

but it's fine on

mohamed abdel-hady

unread,
Feb 2, 2019, 3:47:40 AM2/2/19
to pagespeed-insights-discuss
Yes,
I have noticed this couple of hours ago, and it was driving me nuts,
I tested my site on several testing websites and they all return awesome results, whereas google lighthouse scores have suddenly dropped significantly, and upon following the reason why, turns out that google has stopped retrieving files compressed, god knows why

Sam Maurus

unread,
Feb 2, 2019, 5:25:36 AM2/2/19
to pagespeed-insights-discuss
Same here. We just noticed this this morning. We confirmed the same problems on our site as well as with a test of https://stackoverflow.com using the JavaScript API.

Reproduction:
* Open Chrome and go to https://stackoverflow.com
* Open devtools network tab and see that the JS resource https://ajax.googleapis.com/ajax/libs/jquery/1.12.4/jquery.min.js had a network payload size of 33.2KB (i.e. compressed).
* Switch to console tab and enter:


var start = function() {
  // 2. Initialize the JavaScript client library.
  gapi.client.init({
    'apiKey': "<TODO VALID API KEY>"    
  }).then(function() {
    // 3. Initialize and make the API request.
    return gapi.client.request({
    })
  }).then(function(response) {
    console.log(response.result);
  }, function(reason) {
    console.log('Error: ' + reason.result.error.message);
  });
};

gapi.load('client', start);

* Wait for the console to log the results
* Navigate to lighthouseresult.audits.total-byte-weight.details.items in the results and notice that the same jQuery resource "totalBytes" is 97.7KB (i.e. no compression)

StackOverflow has no such error when running lighthouse from within the browser.

Interestingly, we logged the user agent on our server and saw that the accept-encoding header was set sensibly (gzip.deflate,br), so I'm really not sure why it's not using the compressed result returned from our server. Perhaps it's measuring the uncompressed size, which would indeed be a bug.

It seems indeed that a recent update has introduced a regression in this respect.

rajesh singh

unread,
Feb 3, 2019, 10:46:40 PM2/3/19
to pagespeed-insights-discuss
I am managing a few Divi template based WordPress site and there score has come down crashing to 5 or 20 from earlier in 80 to 90

and the only difference that can be seen is the CSS file are now shown as uncompressed in page speed score (because of this page speed is counting extra load time of 3 to 5 seconds). Anyone knows if this is going to change or it will remain like this

Marcus Williams

unread,
Feb 4, 2019, 10:56:44 AM2/4/19
to pagespeed-insights-discuss
Yes I'm seeing this too, all of a sudden lighthouse no longer recognises gzipped assets.

devquipèteuncable

unread,
Feb 4, 2019, 12:14:05 PM2/4/19
to pagespeed-insights-discuss
I didn't sleep for 24h looking for what's happen with my site.
I've contacted all my plugin support to ask what's happen, if they heard about similar case.
I've almost broke my site and did few FTP and DB backup....

And then ! I see this groupe hahahahahahah

So what's the hell is going on?!!

Is it a bug or do we have to clean the CSS of our sites?!
....

In my case, I use to write some "preload" font in the header, and I've noticed the requesting time has increased a lot.
That's why I think this is a bug.. (hope)

David Gargiulo

unread,
Feb 4, 2019, 2:04:20 PM2/4/19
to pagespeed-insights-discuss
Update: Lighthouse may have identified the issue and they are working on a fix

Michael Smyth

unread,
Feb 4, 2019, 6:26:28 PM2/4/19
to pagespeed-insights-discuss
Looks like google are rolling out the 'bug fix' now, our scores are returning to normal.

Elizabeth Sweeny

unread,
Feb 4, 2019, 8:25:18 PM2/4/19
to pagespeed-insights-discuss
Hello folks,

We really appreciate your patience with this issue - we identified an incorrect network payload size bug that caused incorrect calculations of the performance metrics and ultimately a lower score. As of 5pm PST today, the bug was completely addressed via a new production rollout.

While this will restore Performance scores for the mobile profile to previous values, scores for desktop were expected to drop and will remain lower than previous, due to the 4.1.0 upgrade detailed here: Lighthouse 4.1 Release Notes

Thank you so much again, and if you have further questions please do feel free to reach out.

Best,
Elizabeth

Michael Smyth

unread,
Feb 4, 2019, 10:39:19 PM2/4/19
to pagespeed-insights-discuss
Thank you for the detailed update Elizabeth.

Raul Peixoto

unread,
Feb 5, 2019, 1:07:34 AM2/5/19
to pagespeed-insights-discuss

These are all great news, thank you for the detailed update.

There is however one more bug for the desktop tool, which hasn't been addressed and it's relatively important and reported here:

If that's something you could possibly fix, that would be great.

Thanks

ATC company

unread,
Feb 5, 2019, 2:48:48 AM2/5/19
to pagespeed-ins...@googlegroups.com
Hi,

still not fixed !!

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pagespeed-insights-discuss/a30ade91-29d8-4ad3-b2a0-e79c37451aff%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Raul Peixoto

unread,
Feb 5, 2019, 2:52:12 AM2/5/19
to pagespeed-insights-discuss
yes it is, but you can post details to refute that it's not fixed.

If your score doesn't go up, it may be something else on your site, or you were not affected by this bug anyway.

To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-discuss+unsub...@googlegroups.com.

ATC company

unread,
Feb 5, 2019, 2:54:40 AM2/5/19
to pagespeed-ins...@googlegroups.com
Hi,, no I have two websites each one was scored 98 , 95% !! now it's 80, 70!!!!
i didn't do anything in the sites,, same as before.

To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pagespeed-insights-discuss/e621236d-6fe4-4f88-b310-6ce24a5798ee%40googlegroups.com.

Raul Peixoto

unread,
Feb 5, 2019, 2:58:02 AM2/5/19
to pagespeed-insights-discuss

That could be true, but pagespeed and google keeps changing the requirements.
That means, your score can be great today and poor tomorrow.
It's the same as ranking your site on google, it can rank well today and tomorrow an algorithm update brings it down.

The bug reported here was for compression not being considered after the update, which brough most sites to scores from 0 to 10.
If you score 70, that means the requirements got tighter and you need to optimize the site further.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-discuss+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-discuss+unsub...@googlegroups.com.

ATC company

unread,
Feb 5, 2019, 3:08:39 AM2/5/19
to pagespeed-ins...@googlegroups.com
:( I spent lots of time optimize my websites :( ... and I tested them for around three moth every day and the score was great ... why they change it!!! it's not fair !!!! 

To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pagespeed-insights-discuss/ff96448f-c8f3-4f2b-a6f5-83fbe5b812ab%40googlegroups.com.

David D

unread,
Feb 6, 2019, 4:52:47 AM2/6/19
to pagespeed-insights-discuss
My Mobile Score is still not back to its normal state - It did increase after you fixed the compression bug, but it has dropped about 5 points than what it used to be.

From the insight report, I keep seeing 3 main warnings, that have never been displayed before few days ago (didn't change anything in site):

First CPU Idle - Orange range - used to be green
Time to Interactive - Orange range - used to be green
and Minimize Main thread work - Orange as well - did not show at all before.

My site keeps loading at the same speed at all other insight tools such as gtmetrix and pingdom.

Therefore, It seems like there was *another* bug, which affects mobile, and the issue was not only compression analysis bug.

Please look into this Google Team.

(For Desktop the score seems fine again.)

sougata ghosh

unread,
Feb 6, 2019, 8:11:06 AM2/6/19
to pagespeed-ins...@googlegroups.com
at first clear ur cache....

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.

David D

unread,
Feb 6, 2019, 8:52:19 AM2/6/19
to pagespeed-insights-discuss
It has nothing to do with my own cache. Google's machine is the client who's making the test. 
Either way I'm making the test on the page with an added parameter such as page.php?test-1550PM and it's same low scores compared to earlier in January.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-discuss+unsub...@googlegroups.com.

Raul Peixoto

unread,
Feb 7, 2019, 12:30:42 AM2/7/19
to pagespeed-insights-discuss

There's no bug, have you read the release notes?

What you are seeing are new changes on the PSI tool, so basically you need to optimize your site even further if you want to have better scores.
The PSI tool is never meant to be static, it keeps changing with every update.
It's the same as the rankings on google search, it changes when they update their algorithm and you cannot go and complaint that before it was ranking higher...
Or perhaps you can, but you are just going to be ignored by google because they are the ones to decide the requirements at any time they want.

As for adding a query to bypass cache... that's just going to measure your score without page cache.
That's your real score without cache... the higher one you see, is obviously with the page cache on your server.
That's why people use cache, to make the site loads faster.

David D

unread,
Feb 7, 2019, 5:00:35 AM2/7/19
to pagespeed-insights-discuss
I read the release notes and they only mention that desktop scores are expected to drop, not mobile. In my case, Desktop score remained identical, but mobile has dropped!
Reply all
Reply to author
Forward
0 new messages