SiteSpeed: investigate Project File Source test failure
Latest SiteSpeed pipeline against 10k failed on web_project_file_source
test - the page is not being loaded and crushes the Chrome.
Seeing the same behaviour on a local machine when trying to navigate http://10k.testbed.gitlab.net/gpt/large_projects/gitlabhq1/blob/master/fixtures/emojis/index.json or https://staging.gitlab.com/gpt/large_projects/gitlabhq1/blob/master/fixtures/emojis/index.json.
Worth noting that the same page didn't break the SiteSpeed pipeline last week, so the issue requires investigation.
UPDATE
A new request appeared comparing to the last week results on Friday - http://10k.testbed.gitlab.net/assets/webpack/chrome_84_icon_fix.ef9c247c.chunk.js
- it takes roughly 26676 ms
of the CPU time. This new request tentatively appeared when gitlab-org/gitlab!38304 (merged) was merged late Friday night - it adds a new library svg4everybody
and it forcefully includes the icon fragments into the DOM and thus circumventing the bug in Chrome 84. More details in #312 (comment 390538757).
We also see a significant degradation in web_project_file_blame
- TBT 55 sec
vs 26.7 sec
last week:
NAME | FCP (ms) | LCP (ms) | TBT (ms) | SI (ms) | LVC (ms) | TFR SIZE (kb) | SCORE | RESULT
--------------------------------------|----------|------------------|------------------|---------|----------|---------------|-------|-------
web_project_file_blame | 4173 | ✓ 4173 (<5000) | ✘ 55303 (<32500) | 4980 | 61350 | 936.0 | 87 | FAILED
web_project_file_source | - | - | - | - | - | - | - | FAILED
Last week:
NAME | FCP (ms) | LCP (ms) | TBT (ms) | SI (ms) | LVC (ms) | TFR SIZE (kb) | SCORE | RESULT
--------------------------------------|----------|------------------|------------------|---------|----------|---------------|-------|-------
web_project_file_blame | 4825 | ✘ 4825 (<4500) | ✓ 26748 (<32500) | 5242 | 34300 | 935.8 | 87 | FAILED
web_project_file_source | 1013 | ✓ 22507 (<27500) | ✓ 16245 (<20000) | 2581 | 23767 | 878.3 | 85 | Passed