1903104 - Banking site (www.partnersfcu.org) broken since 126.0.1 (due to removal of -moz-transition)

Status: NEW

(Web Compatibility :: Site Reports, defect, P1)

Description

[lohphat]

Attachment: Screenshot 2024-06-17 at 1.00.45 PM_Redacted.png

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:127.0) Gecko/20100101 Firefox/127.0

Steps to reproduce:

Site: https://www.partnersfcu.org/

I can login to my bank account but can't pull up detail. I get "Loading..." pop up and the site hangs when I try to access account transaction details.

Site works on Chrome and Edge.

I've disabled Adblock and tracking protections for site.

Actual results:

Get spinning "Loading..." pop up, no progress. Page then becomes unresponsive. Reloading page returns to login page.

Expected results:

"Loading..." pop up should then give way and display transaction history.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'Core::Networking' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[lohphat]

If I can get a pre-126 installer I might be able to get you some packet captures to compare if that helps.

Comment 3

[Randell Jesup [:jesup] (needinfo me)]

You can get pre-126 installers here:
https://archive.mozilla.org/pub/firefox/releases/

You can also use mozregression which can identify the exact change that broke your bank: https://mozilla.github.io/mozregression/
For a 'good' date, I'd suggest using 2024/1/1, since I'm certain that should have worked. Note that mozregression will work with Nightly and autoland builds, which lets us figure out what checkin caused it.

Normally I'd ask for network logs, but those might have sensitive information.

Thanks!! We'd love to track this down and solve it

Comment 4

[lohphat]

The regression tool didn't work. It actually made it worse. Now my accounts won't load at all on my bank account dashboard.

I've nuked my profile and rebuilt from scratch -- no joy.

If I install the 124.0.2 installer my bank site works again. but then it re downloads 127 so I can't exit the browser or it will auto update.

Comment 5

[lohphat]

Attachment: mozregression output

Comment 6

[lohphat]

Scratch that I had to go back further than I thought. See output attachment for full log.

Here's the summary:

...
2024-06-18T00:28:20.026000: INFO : Narrowed integration regression window from [440bcf1d, de35cb5d] (3 builds) to [ffdf085e, de35cb5d] (2 builds) (~1 steps left)
2024-06-18T00:28:20.042000: DEBUG : Starting merge handling...
2024-06-18T00:28:20.042000: DEBUG : Using url: https://hg.mozilla.org/integration/autoland/json-pushes?changeset=de35cb5de97565935b503eade2990b5c385c7293&full=1
2024-06-18T00:28:20.042000: DEBUG : redo: attempt 1/3
2024-06-18T00:28:20.042000: DEBUG : redo: retry: calling _default_get with args: ('https://hg.mozilla.org/integration/autoland/json-pushes?changeset=de35cb5de97565935b503eade2990b5c385c7293&full=1',), kwargs: {}, attempt #1
2024-06-18T00:28:20.045000: DEBUG : urllib3.connectionpool: Resetting dropped connection: hg.mozilla.org
2024-06-18T00:28:20.941000: DEBUG : urllib3.connectionpool: https://hg.mozilla.org:443 "GET /integration/autoland/json-pushes?changeset=de35cb5de97565935b503eade2990b5c385c7293&full=1 HTTP/1.1" 200 None
2024-06-18T00:28:20.994000: DEBUG : Found commit message:
Bug 1865332 - Expose MozTransform in CSSStyleDeclaration unconditionally. r=jwatt

Shipping CSS zoom requires to remove -moz-transform from stylesheets.
That's generally fine, however we've seen a bunch of regressions due to
sites doing broken feature detection in JS (see bugs mentioned in the
patch).

It seems like exposing MozTransform in CSSStyleDeclaration is a
requirement, for some large amount of pages. The hypothesis here is that
setting both MozTransform and zoom using JS is uncommon.

We could expose it conditionally, with some sort of reverse origin trial
or what not but, given the amount of regressions only on Nightly and
that this is harmless if you do proper feature detection, I think I'd
rather expose it unconditionally, rather than playing whack-a-mole
fixing pages after the fact. We could play whack-a-mole with pages that
use zoom and -moz-transform, if needed in the future, which seems like a
safer path forward.

Note that this change currently only affects behavior on Nightly,
because on other release channels -moz-transform is enabled right now.

Differential Revision: https://phabricator.services.mozilla.com/D194006

2024-06-18T00:28:20.994000: DEBUG : Did not find a branch, checking all integration branches
2024-06-18T00:28:20.994000: INFO : The bisection is done.
2024-06-18T00:28:20.998000: INFO : Stopped

Comment 7

[lohphat]

Attachment: Fill 6.1.0 mozregression log

I rant it again but with the release 6.1.0 mozregression and got this:

...
2024-06-18T01:02:13.411000: INFO : platform_version: 121.0a1
2024-06-18T01:02:59.806000: INFO : Narrowed integration regression window from [440bcf1d, de35cb5d] (3 builds) to [440bcf1d, ffdf085e] (2 builds) (~1 steps left)
2024-06-18T01:02:59.830000: DEBUG : Starting merge handling...
2024-06-18T01:02:59.830000: DEBUG : Using url: https://hg.mozilla.org/integration/autoland/json-pushes?changeset=ffdf085e5e3ca95d10e929d78a0b119dc5fc2cbc&full=1
2024-06-18T01:02:59.830000: DEBUG : redo: attempt 1/3
2024-06-18T01:02:59.830000: DEBUG : redo: retry: calling _default_get with args: ('https://hg.mozilla.org/integration/autoland/json-pushes?changeset=ffdf085e5e3ca95d10e929d78a0b119dc5fc2cbc&full=1',), kwargs: {}, attempt #1
2024-06-18T01:02:59.838000: DEBUG : urllib3.connectionpool: Resetting dropped connection: hg.mozilla.org
2024-06-18T01:03:00.763000: DEBUG : urllib3.connectionpool: https://hg.mozilla.org:443 "GET /integration/autoland/json-pushes?changeset=ffdf085e5e3ca95d10e929d78a0b119dc5fc2cbc&full=1 HTTP/1.1" 200 None
2024-06-18T01:03:00.815000: DEBUG : Found commit message:
Bug 1861730 - Remove transition prefixes from nightly too. r=jwatt

Code like the one from comment 7 seems reasonable. Let's try to disable
prefixed transitions along with prefixed transforms in order to minimize
the compat fallout from CSS zoom.

While I was at it I also fixed the prefixed animation entries in the
property database and crashtests. But those aliases remain enabled.

Differential Revision: https://phabricator.services.mozilla.com/D192129

2024-06-18T01:03:00.815000: DEBUG : Did not find a branch, checking all integration branches
2024-06-18T01:03:00.816000: INFO : The bisection is done.
2024-06-18T01:03:00.817000: INFO : Stopped

Comment 8

[lohphat]

This MAY be related to another bug (bug 1584402) with similar behavior I'm seeing on another site (no registration needed) where expanding a list into more detail fails to load.

Comment 9

[lohphat]

Whoops, wrong bug. It's actually bug 1904079

Comment 10

[Randell Jesup [:jesup] (needinfo me)]

I'm a little confused; where does this bug stand?

It's very odd that trying older versions would affect your ability to load the site - a log of that would be useful too. Perhaps the bank thinks your IP is being used to try to do something nefarious and they've blocked you. Still seems odd; and if 124 works then that wouldn't fit.
You should be able to try 125 and 126 from that archive.
You may also be able to use mozregression, just giving it a longer back starting point (maybe starting 6 months ago).
You could submit logs (which might have sensitive info, note, if you have to sign in) to [e-mail]. Only do this if you're comfortable with our engineers seeing the logs. We do delete them after using them to investigate. Logs taken without logging into the site are probably free of problematic info.

Thanks!

Comment 11

[lohphat]

The log output from above WAS from mozregression.

Comment 12

[Randell Jesup [:jesup] (needinfo me)]

Sorry, one log was with a comment that said

Code like the one from comment 7 seems reasonable. Let's try to disable
prefixed transitions along with prefixed transforms in order to minimize
the compat fallout from CSS zoom.

While I was at it I also fixed the prefixed animation entries in the
property database and crashtests. But those aliases remain enabled.

and the other log was with

Bug 1865332 - Expose MozTransform in CSSStyleDeclaration unconditionally. r=jwatt

Shipping CSS zoom requires to remove -moz-transform from stylesheets.
That's generally fine, however we've seen a bunch of regressions due to
sites doing broken feature detection in JS (see bugs mentioned in the
patch).

So I figured those were meant for a different bug.

The regression range from one of those was this:

	78ec85b2b0a0a1c1e6715d4ac344fd86dfbefcf6	Joel Maher — Bug 1864755 - Adjust mozinfo names/values to be more streamlined and accurate. r=aryx
	0b9eff3819bcc1ea3784903a491fcdefd1c90f37	Jonathan Kew — Bug 1865482 - Make CSS :lang() matching more correctly follow the BCP4647 Extended Filtering algorithm. r=layout-reviewers,emilio
	ea1dc493f7d2eda5baf653701a75bce9f9a31f79	Valentin Gosu — Bug 1760121 - Fix test_trr.js racing get-ttl with dns.clearCache r=necko-reviewers,kershaw
	b7ce1a446e6abacad48a7e958371b42c7433d539	Zhao Jiazhong — Bug 1865570 - [loong64][mips64] Update a function prototype in simulator. r=jseward DONTBUILD
	7626d9d131fb4daff7deedc79b7aa27eef41aa25	Kershaw Chang — Bug 1863995 - Make webtransport datagram test more robust, r=necko-reviewers,valentin
	c1b9ee189d81255b2e573a0d88604cf447af8a4f	Manuel Bucher — Bug 1863995 - neqo v0.6.7 r=necko-reviewers,kershaw
	8d79f4aa58a79c3c19369e71ee6c6a8094c68c2d	Sandor Molnar — Backed out 4 changesets (bug 1651522) for causing dt failures on devtools/shared/commands/target/tests/browser_target_command_service_workers_navigation.js CLOSED TREE
	9517987e19427c2900046f62ca816d243d1f1403	Greg Mierzwinski — Bug 1843851 - Add a tp6 benchmark test to raptor. r=perftest-reviewers,kshampur
	947257872d2a647ed7a31c8b0d1d10a9685fa223	Greg Mierzwinski — Bug 1843851 - Improve support for Raptor Python support files in result/output parsing. r=perftest-reviewers,kshampur
	9b797f214f1999e79623b4afc161911d82bbf0b3	Greg Mierzwinski — Bug 1861663 - Increase unity-webgl tier, and max runtime. r=perftest-reviewers,afinder DONTBUILD
	38489d98c5bcdeb3759137e96ed433b4ced7b298	Olli Pettay — Bug 1861848, update browser_gc_schedule.js to test background tabs only, r=pbone
	86eed183b6802abc57828e9c1baa94e42e53ee03	Olli Pettay — Bug 1861848 - Let foreground content processes run idle tasks without consulting the parent process, r=farre,pbone
	de35cb5de97565935b503eade2990b5c385c7293	Emilio Cobos Álvarez — Bug 1865332 - Expose MozTransform in CSSStyleDeclaration unconditionally. r=jwatt
	ffdf085e5e3ca95d10e929d78a0b119dc5fc2cbc	Emilio Cobos Álvarez — Bug 1861730 - Remove transition prefixes from nightly too. r=jwatt

So my understanding is this bug is apparently due to the moz-transform landing. (bug 1865332) - reassigning

Comment 13

[lohphat]

Those two mozregressions were for the same problem but run on two different version of mozregression. The first was from 6.22-beta and the other from 6.10 release.

Comment 14

[Emilio Cobos Álvarez (:emilio)]

Hmm, sorry for asking, but without a repro it's hard to pin this down. Few questions:

• If you go to about:config, and flip layout.css.prefixes.transforms to true, then reload, does it work?
• If not, what if you do the same but with layout.css.prefixes.transitions?

This would allow to confirm whether this comes from bug 1865332 or bug 1861730.

That said, this is probably the site doing some broken prefix detection (something like if (navigation.userAgent.includes("Firefox")) { /* assume -moz- prefix exists } */...

It'd be good if you could attach the scripts of the page in a zipfile as shown by the debugger or something, but since that might be a bit time-consuming, it'd be nice to at least confirm if loading the page with a Chrome or Safari user-agent works... Would you be able to do that? The right way to do that is to, in about:config again, create a pref named general.useragent.override as a String, with a value like this maybe, which matches Chrome release: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36.

Thanks!

Comment 15

[lohphat]

Setting layout.css.prefixes.transitions=true fixed it!

Comment 16

[Emilio Cobos Álvarez (:emilio)]

Thanks, so this is really a regression from bug 1861730.

Does this work with those prefs back to default but with a Chrome UA override? (That would allow us to fix it for this site in particular)

Comment 17

[Emilio Cobos Álvarez (:emilio)]

If not, we need to find a more subtle solution, but it's hard to know which one would that be without being able to reproduce.

Comment 18

[Emilio Cobos Álvarez (:emilio)]

I tried to look for a contact form or something to report this bug to the site (since it seems they're doing something broken), but unfortunately it seems you need to have an account or something? I don't see any contact email or form that I could fill from https://www.partnersfcu.org/contact-us :(

Comment 19

[Daniel Holbert [:dholbert]]

I'm calling their customer-service number from https://www.partnersfcu.org/contact-us (currently on hold) to see if they can give us a way to report this issue to the folks that maintain this credit union's website.

Comment 20

[Daniel Holbert [:dholbert]]

(update: my telephone customer-support agent gave me an email address for a Partners FCU tech-support person, and emilio's going to reach out to them.)

Comment 21

[lohphat]

(In reply to Emilio Cobos Álvarez (:emilio) from comment #16)

Thanks, so this is really a regression from bug 1861730.

Does this work with those prefs back to default but with a Chrome UA override? (That would allow us to fix it for this site in particular)

I don't know what a Chrome UA override is. Treat me like: "Leeloo -- Mooltipass!" level of understanding.

I'm not a dev, I haven't played one on TV in 30 years. ;-)

PS if you don't get the movie reference....Son, I am dissapoint.

Comment 22

[Emilio Cobos Álvarez (:emilio)]

(In reply to Daniel Holbert [:dholbert] from comment #20)

(update: my telephone customer-support agent gave me an email address for a Partners FCU tech-support person, and emilio's going to reach out to them.)

Done (cc'd), hope we hear back :)

(In reply to lohphat from comment #21)

I don't know what a Chrome UA override is. Treat me like: "Leeloo -- Mooltipass!" level of understanding.

Yeah, sorry, an override would be to do the last bit I wrote in comment 14, in particular the:

in about:config again, create a pref named general.useragent.override as a string, with a value like this maybe, which matches Chrome release: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36.

If you could do that (see here for a guide with screenshots), then reload, and confirm it works, that means we had a short-term fix for this site we could apply.

Please revert that when you're done because that could break other sites.

PS if you don't get the movie reference....Son, I am dissapoint.

Sorry! ;_;

Comment 23

[lohphat]

Reverting layout.css.prefixes.transitions back to the default (false) and adding the Chrome UA worked.

I will try this at home on my non-Mac box to see if the same fix works.

(BTW it was The Fifth Element https://www.youtube.com/watch?v=8bF5ft-oOWU)

Comment 24

[lohphat]

This is broken on both Mac and Windows.

Comment 25

[Emilio Cobos Álvarez (:emilio)]

(In reply to lohphat from comment #23)

Reverting layout.css.prefixes.transitions back to the default (false) and adding the Chrome UA worked.

Thanks! Seems we could do an intervention for this site then. Not sure who would be the right person to make that call... we can also wait a few days for their support to get back to us if needed.

I will try this at home on my non-Mac box to see if the same fix works.

Sweet, that'd be great!

(BTW it was The Fifth Element https://www.youtube.com/watch?v=8bF5ft-oOWU)

I was 2 when that film came out, but Daniel explained the pun to me and now I need to watch it ;-)

Comment 26

[lohphat]

(In reply to Emilio Cobos Álvarez (:emilio) from comment #25)

I was 2 when that film came out

I was 2 when "2001: A Space Odyssey" came out!

Lawn...OFF! ;-)