Jonathan Watt [:jwatt] Assignee
	Description • 11 years ago

Bug 635240 is adding various CSS pseudo-elements to allow the various anonymous parts of <input type=number> to be styled by content authors. We've not fixed how that should behave just yet though. We should have reftests before shipping this.

	Jonathan Watt [:jwatt] Assignee
	Comment 1 • 11 years ago

Bug 930010 means that we can't do this for now since we don't have support for running reftests with chrome privs. We could maybe test this using chrome mochitests that use windowsnapshot instead, maybe. I don't plan on trying that until January though, so if anyone wants to have a go feel free.

	Daniel Holbert [:dholbert]
	Comment 2 • 11 years ago

(In reply to Jonathan Watt [:jwatt] from comment #1)
> Bug 930010 means that we can't do this for now since we don't have support
> for running reftests with chrome privs. We could maybe test this using
> chrome mochitests that use windowsnapshot instead, maybe.

(Or perhaps we should add a hidden (i.e. not listed in all.js) pref that makes us turn on "mUnsafeRulesEnabled" even for content... Then we could stick with reftests, and just set the pref in reftest.list?)

	Jonathan Watt [:jwatt] Assignee
	Comment 3 • 11 years ago

Hmm, in the past we've had instances of people being tricked into flipping prefs that make them insecure. Not sure that would be something we'd want to ship.

	Daniel Holbert [:dholbert]
	Comment 4 • 11 years ago

That's why I suggested making it a hidden pref, so that it won't be listed in "about:config" and people wouldn't be able to just flip it.  (and of course we'd also give it a sufficiently-scary-sounding name like e.g "layout.css.dangerousUnsafeRules.enabled")

But yeah, agreed - shipping with it forcibly off (or with the code that reacts to it #ifdeffed out) in release builds would be even safer (though it'd make testing more complicated).

Anyway, I'll stop bikeshedding now. :)

	Jonathan Watt [:jwatt] Assignee
	Comment 5 • 11 years ago

Okay, as long as we're mindful about the risk. :)

	Jonathan Watt [:jwatt] Assignee
	Comment 6 • 10 years ago

Note to self: be sure to address the last sentence in bug 951310 comment 2 when adding tests here.

	Jonathan Watt [:jwatt] Assignee
	Comment 7 • 10 years ago

And the last sentence in bug 951310 comment 3.

	Daniel Holbert [:dholbert]
	Comment 8 • 10 years ago

We also need a reftest or crashtest that has something like
  <input type="number" style="width:auto">
with display:none on the outermost pseudo-element (the "outer wrapper"), per bug 970257 comment 3, top chunk.