regresion range is https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=6d38ad302429c98115c354d643e81987ecec5d3c&tochange=9a26ed658fdc1d14c6153451e7e45f6ec94292fb But I think that this might be regression by bug 1318312. - Step 1. Open twitter.com 2. Post a Tweet - Result tweet textbox is flicker. height of textbox is always changed.

Oh, okay, this is a simple mistake. When we move focus, we keep mCalledByJS true. Therefore, internal selection change causes nested focus move.

Hmm, I tried to write a test to call a method of nsISelection directly. However, I cannot do that since even with QueryInterface or getting from nsISelectionController, a call of method causes running Selection::*JS() (i.e., defined by Selection.webidl). If somebody knows how to write such test, let me know. # I'm going to request to review it without test because of a serious regression on major web service.

Ah, nsRange could have same bug. I'll work on it in follow up bug.

Comment on attachment 8848060 [details] Bug 1347809 Set Selection::mCalledByJS to false before moving focus in NotifySelectionListeners() because non-*JS() methods don't set it to false https://reviewboard.mozilla.org/r/120974/#review123122 ::: commit-message-d57b3:1 (Diff revision 1) > +Bug 1347809 Unmark selection change is caused by JS before moving focus r?smaug I don't understand this commit message

Pushed by masayuki@d-toybox.com: https://hg.mozilla.org/integration/autoland/rev/cfd64c2b4970 Set Selection::mCalledByJS to false before moving focus in NotifySelectionListeners() because non-*JS() methods don't set it to false r=smaug