Screenshots timing out

[twolfson]

In a recent Nightmare script, I was consistently getting timeouts occurring and wound up writing #945 to resolve timeouts not cleaning up after themselves properly. This issue is mostly for @Mr0grog so they can debug issues on triggering frames in the said script. I made a minimal example here:

https://gist.github.com/twolfson/b666177b87ae339f3ab2183c2be77690

Things I noticed:

• Removing the wait(100) will cause it to start passing
• Removing 1 screenshot call will cause it to start passing
• Moving from getbootstrap.com to google.com will cause it to start passing

Best of luck on your debugging!

/cc @Mr0grog

[Mr0grog]

OK! Sorry I let this fall off my radar. #964 reminded me of it—and it turns out this issue and that one are related!

It appears to be possible that the DOM.highlightRect and DOM.hideHighlight debugger calls can happen in rapid enough succession that no frames are generated (both operations happen before generating a frame, so nothing has “changed” visually, and so no frame is rendered).

There also seems to be an issue where the synchronous call to detach the debugger immediately after the DOM.hideHighlight call prevents the hideHighlight operation from ever running—so on the second screenshot, the call to DOM.highlightRect isn’t actually changing anything.

Anyway, I think the solution to both this and #964 is to:

1. Call DOM.highlightRect
2. Wait for a frame
3. Call DOM.hideHighlight
4. Wait for a frame
5. Detach the debugger
6. Dispatch a data event from FrameManager

This will probably be easiest if I wait for #945 to be complete, since it’ll be messing around in the same area of the codebase.

[FredrikNoren]

Experiencing the same issue here

[entrptaher]

Calling screenshot function multiple times resulted in this problem.

Related Issue: #555

[chunsli]

Bumping this up! Experiencing the same issue as well! Thanks guys!

[lokidokicoki]

bump! same issue

[ucjonathan]

bump, same issue

[Mr0grog]

Y'all should probably be bumping #945 since this depends on that.

[Mr0grog]

@casesandberg would you like to see a PR for this now that #945 is merged, or is this pretty much being left alone now that you are moving towards headless Chrome?

[Howardwlo]

@Mr0grog I'm running into this issue and I'd love to get the PR :). What's this about headless Chrome?

[Mr0grog]

The Segment folks are rewriting this library to use Chrome (headless) and not Electron, so the way a lot of stuff like this that heavily depends on the API for controlling the browser will have to be completely different.

[Howardwlo]

@Mr0grog Oohhh. Are there updates about the switch to Chrome somewhere?

[Mr0grog]

I think it’s on the v3 branch. I don’t know any details beyond what I’ve already stated, which was what @casesandberg told me in a reply on another issue (lost track of it now).

[matthewmueller]

Hey folks, we're going to stick with electron for now since puppeteer is already a great option for headless chrome.

I'm assuming the merged PR resolved this issue, so I'm going to close. If it's still a problem, please reopen with what's still not working – thanks!

[Mr0grog]

If the merged PR you are referencing is #945, it does not solve this issue (though it was blocking the solution/the work for this issue).

[c-vetter]

@matthewmueller @Mr0grog
Is this fixed, or should it be reopened?

[Mr0grog]

No change from my previous comment—unless the code in frame-manager.js has been changed (and it does not appear to have been), then no, this is not fixed.