1412340 - firefox 56 javascript submit form: action attribute behavior changes

Status: RESOLVED WONTFIX

(Core :: DOM: Core & HTML, defect, P3)

Description

[Marc Foret]

Attachment: index.html

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:56.0) Gecko/20100101 Firefox/56.0
Build ID: 20171003100843

Steps to reproduce:

When the "action" attribute is not defined for a form. 
Previously browser returns an empty string.
Now Firefox 56 reply with the url of the page, all other browsers return an empty string.

Any Web Application replying on this behavior is possible broken in Firefox 56.

Live check on http://marc.foret.free.fr/bug_form_action 


Actual results:

form.action= "http://marc.foret.free.fr/bug_form_action";


Expected results:

form.action = ""; //

Comment 1

[Kohei Yoshino [:kohei]]

Do you know if this is a recent regression in Firefox?

Comment 2

[Marc Foret]

yes only in 56 and 57beta.

I have discover it yesterday at the client. The web application was working since a few years. 
I have only download Firefox 52ESR, 54 and 55 in order to test and they don't have the issue.

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

It would be great to have regression range here.
(http://mozilla.github.io/mozregression/ may help)

Comment 4

[Alice0775 White]

Regression window:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b8783d6fffdbebed01455743185cea68784d5367&tochange=d43779e278d2e4d3e21dba2fcb585a3bf4b1288e

Regressed by: 
	8f7663ac2068	Jessica Jong — Bug 1366361 - Part 2: Update reflectURL() and web platform tests. r=smaug
	1c1a86e1296b	Jessica Jong — Bug 1366361 - Part 1: .action/formAction should return the document's URL if @action/formaction is missing or empty. r=smaug

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

Thanks Alice!

Comment 6

[Andrew Overholt [:overholt]]

So did we change our behaviour to match the spec and that broke sites? Is there anything we can/should do to warn here?

Comment 7

[Andrew Overholt [:overholt]]

Oh, I hadn't seen https://bugzilla.mozilla.org/show_bug.cgi?id=1366361#c17 so maybe we should move this to TE?

Comment 8

[Jessica Jong [:jessica] (inactive)]

Bug 1408658 reports the problem on a real website, this one is just reporting the behavior on a test page. So maybe we should close this one and move Bug 1408658 to TE?

Comment 9

[Marc Foret]

For me you can close this issue.

Just explain my case that is also real case: 
 - we have developed a web server embedded into a Constant current regulator(equipment that control light on the airfield).
 - web server is running on Micro-controller (limited resources).
 - we switch to aweb browser interface for maintenance/configuration instead of developing a dedicated desktop program.
    - for a lot of good reasons. one of the reasons, you only need a standard browser to configure....

And now after the long validation/certification process of the production release. Suddenly (after release 56), our recommended browser(Firefox) is not working anymore.
- the problem was discovered on site in the airport...
I never have expected, that kind of issue. ( even it is right, all browser don't implemented the spec correctly).

This kind of changes must be documented as behavior changes. and distribute to community as potential breaking changes.

Must be also shared with all WEB standard communities...
What is the position of the other vendors on this behavior changes ?

Even if the fix is already existed for the web server, but it is impossible/difficult to push to the installed based ((re-validation process, customer relation, airport access).

If it was a Internet web server, i didn't reported the issue and patch the web server directly and find a way to rapidly release to production. In my case is closed to be impossible.

Comment 10

[Jessica Jong [:jessica] (inactive)]

Hi Marc, 

Sorry about the problem caused. I agree that the change should be documented and distributed to community before release. We'll be more careful in the future when making this kind of changes. Thanks for the report and feedback.

Comment 11

[Kohei Yoshino [:kohei]]

I could post a site compatibility note earlier to fxsitecompat.com if Bug 1366361 had the dev-doc-needed keyword. Anyway, it should be safe to close this bug.