User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:55.0) Gecko/20100101 Firefox/55.0 Build ID: 20170601030206 Steps to reproduce: 1. Have a page loaded in the current tab 2. Hit ^t to open a new tab 3. Hit ^z Actual results: ^z does nothing. Expected results: Before, hitting ^z would fill the address bar with the URL from the previous tab. Using mozregression, someone else narrowed the change down to: INFO: Last good revision: 92c5898618c21d80d8cccd49bc08ec5fc8924a45 INFO: First bad revision: 029d3143f52d17a218cd7dcb33ec3003fe5a8d9d INFO: Pushlog: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=92c5898618c21d80d8cccd49bc08ec5fc8924a45&tochange=029d3143f52d17a218cd7dcb33ec3003fe5a8d9d INFO: Looks like the following bug has the changes which introduced the regression: https://bugzilla.mozilla.org/show_bug.cgi?id=1358025 I don't actually know if the previous behavior was intended, but I did use it.

Makoto, could you check why undo is disabled in this case.

(In reply to Loic from comment #1) > Makoto, could you check why undo is disabled in this case. As textbox, this is expected behavior (all browsers are same behavior). Because setting to about:blank etc isn't user operation (it means that user doesn't input it). If comment #0's step is expected as product design / spec, we can use setUserInput instead of input.value setter. setUserInput works on chrome and urlbar is on chrome.

(In reply to Makoto Kato [:m_kato] from comment #2) > If comment #0's step is expected as product design / spec, we can use > setUserInput instead of input.value setter. setUserInput works on chrome > and urlbar is on chrome. Who would make the call on that?

(In reply to Julien Cristau [:jcristau] from comment #3) > Who would make the call on that? Marco, who is owner of location bar spec?

I think in general comment 2 is right, so this is a wontfix. Opening a new tab should clear the transaction history, since the location bar appears inside the tab, so the context is a different one. That said, if there is some common use-case that may have be widespread and is now broken, we're open to evaluate those and issue specific fixed for them. But we need specific bugs for the use-case to be able to evaluate. Regarding this specific case "open a new tab and ^Z to go the url of the previous tab", it sounds exotic and breaks the above assumption about separate contexts. Thus is imo a wontfix. you can ctrl+drag a tab to create a dupe of it. I agree it's not great from the keyboard, but ^C,^T,^V is still usable, we could eventually evaluate Ctrl+Enter doing the same when canonization (adding www.*.com) doesn't apply to the value.

looks like it's *fixed* by bug 1386222 and now history is shared.