tests: use --headless=new for all smoketests #14419
Conversation
paulirish
changed the title
| @@ -61,6 +61,8 @@ async function internalRun(url, tmpPath, configJson, options) { | |||
| `${url}`, | |||
| `--output-path=${outputPath}`, | |||
| '--output=json', | |||
| // native headless, as full headless has several behavior differences that break tests. http://go/rooxt | |||
| '--chrome-flags="--headless=chrome"', | |||
There was a problem hiding this comment.
Looking at my (stale) branch doing the same, here are some things to consider:
- there is also
lighthouse-runners/bundle.jsto update - headless-chrome.md could mention this new mode. what I had:
+There is also the new
--headless=chromeoption, which includes functionality that
+was explicitly omitted from the original headless browser.
+CLI (xvfb)
Last time I tried this, there were issues re: the protocol hanging on screenshots, IIRC.
There was a problem hiding this comment.
SG. also i wanted to experiment with using this in CI and getting rid of xvfb, which should (in theory) work.
but i'll do this part in a followup as i expect it'll be slightly more involved
ill address your two bullets in this one
connorjclark
changed the title
done. also. i suspect theres still failures here. I'm gonna pick this PR back up post 10.0 probably. |
adamraine
changed the title
types/internal/smokehouse.d.ts
Outdated
| /** If true, performs extra logging from the test runs. */ | ||
| isDebug?: boolean; | ||
| /** Launch Chrome in typical desktop headful mode, rather than our default of `--headless=new`. */ | ||
| forceHeadful?: boolean; |
There was a problem hiding this comment.
nit: to match puppeteer and be a bit more straight-forward, could this be headless?: boolean (which defaults to true when not specified)?
The cli invocation for not headless would then be --no-headless (I believe yargs handles that for us, if we define headless to be a boolean flag)
There was a problem hiding this comment.
Edit: didn't see your edit
Would that still be --force-headful on the CLI though? If that's the way we are naming the flag I would prefer to just keep the option as forceHeadful throughout.
I've wanted to use headless for CLI smoketests to avoid windows popping up and stealing focus. But as we know, some test assertions fail.
The
--headless=chromeflag provides something much more like Chrome than what the--headlessflag launches. And it now has support across windows/linux/mac.edit: feb 2023: the flag was renamed to --headless-new. GoogleChrome/chrome-launcher#290
🔒 http://go/rooxt and http://go/wkigd