Every non-recognized input type falls back to type="text", so we now rather exclude the types that we don't want instead of trying to be comprehensive as to what types we want. Per <http://hg.mozilla.org/mozilla-central/annotate/6f18a6d94920/content/html/content/src/nsHTMLInputElement.cpp#l1968>, that's the following types to ignore: BUTTON, HIDDEN, RESET, IMAGE, PASSWORD and SUBMIT.

Comment on attachment 339759 [details] [diff] [review] modified XPath query and tests r=me, thanks

backed out TEST-UNEXPECTED-FAIL | chrome://mochikit/content/browser/browser/components/sessionstore/test/browser/browser_456342.js | Didn't save text for ignored field types - Got 3, expected 0

Since 1.9.1, XPath considers attributes to be case sensitive, so we have to lower-case them first. My bad for testing this only on 1.9.0.

Simon or Dao, one question beside. In case of type="search" is it a regression by switching from type="timed" or was its content even not saved?

(In reply to comment #6) You're confusing XUL textboxes (for which there's a deprecated type="timed") with HTML input elements (for which type="search" is currently an inofficial, Webkit introduced extension). If this bug is a regression (which it actually is) then from bug 346337 which restricted too much what is considered to be restorable.

Alright, thanks for clarification Simon.

Pushed to mozilla-central: <http://hg.mozilla.org/mozilla-central/rev/ac1e09aa72ab>

To verify, load this bug's URL, enter some text and then duplicate the tab (e.g. through Ctrl+dragging the tab).