The web page https://www.wpc.ncep.noaa.gov/threats/threats.php stopped displaying the map correctly. See the attached screenshot.

Mozregression shows:

6:32.92 INFO: Last good revision: afe375a5bb0a870f7283fc330c2dc8c8bd2ee965
6:32.92 INFO: First bad revision: 793a305aa0628613e6681be3a57c4b089fdecf4c
6:32.92 INFO: Pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=afe375a5bb0a870f7283fc330c2dc8c8bd2ee965&tochange=793a305aa0628613e6681be3a57c4b089fdecf4c

Patch author — Commit message:
Emilio Cobos Álvarez — Bug 1855763 - Disable -moz-transform and enable zoom on Nightly. r=zcorpan

Set release status flags based on info from the regressing bug 1855763

:emilio, since you are the author of the regressor, bug 1855763, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

So this site assumes that, if we're Firefox, we support the -moz-transform prefix...

var e = d("ff");
// ...
return {
  names: {
    transition: k && "-webkit-transition" || e && "MozTransition" || l && "OTransition" || f && "msTransition" || "transition",
    transform: k && "-webkit-transform" || e && "MozTransform" || l && "OTransform" || f && "msTransform" || "transform",
    transformName: k &&
	"-webkit-transform" || e && "-moz-transform" || l && "-o-transform" || f && "-ms-transform" || "transform",
    origin: k && "-webkit-transform-origin" || e && "MozTransformOrigin" || l && "OTransformOrigin" || f && "msTransformOrigin" || "transformOrigin",
    endEvent: k && "webkitTransitionEnd" || e && "transitionend" || l && "oTransitionEnd" || f && "MSTransitionEnd" || "transitionend"
  },

That code seems to come from esri, which is kind of a pain because that means that we can't do a simple per-site intervention.

Maybe we can do outreach for them to get this fixed, assuming this is isolated enough?

This is nightly-only for now.

|| "transform"

Why doesn't the fallback to unprefixed trandform work?

Because they're not checking whether the browser supports -moz-transform, they're doing something like:

let name = (isWebKit && "-webkit-transform") || (isGecko && "-moz-transform") || ... || "transform";

So isGecko will always end up in the -moz-transform case.

Good times. They should feature check instead. Their map library is often self-hosted, so this issue might linger for a long time even if fixed upstream.

I'll see if I can find a contact for esri.

In fact it seems the problematic code is no longer there in the latest version of ArcGIS (4.28, possibly any 4.x):

https://developers.arcgis.com/javascript/latest/
https://js.arcgis.com/4.28/init.js

...but is still there in the latest "legacy" ArcGIS (3.45, possibly any 3.xx - the page for this bug https://www.wpc.ncep.noaa.gov/threats/threats.php uses 3.2.2):

https://developers.arcgis.com/javascript/3/
https://js.arcgis.com/3.45/init.js

I don't know if they're keen on making changes to the legacy code. But maybe it's possible to target an intervention for scripts in https://js.arcgis.com/3.xx/init.js?

(In reply to Simon Pieters [:zcorpan] from comment #6)

I'll see if I can find a contact for esri.

Email sent.

This should be fixed by bug 1865332.