Node::canStartSelection() can return false on the host node of the selectable input text field for some special cases.
Created attachment 150620 [details] Patch
Comment on attachment 150620 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=150620&action=review could you confirm the original problem fixed by the code you are changing does not regress? > Source/WebKit/blackberry/ChangeLog:6 > + Reviewed by Antonio Gomes. I have not reviewed it :) Tools can fill it up for you
Created attachment 150622 [details] Patch
Created attachment 150631 [details] Patch
Comment on attachment 150631 [details] Patch LGTM.
I can confirm the original problem fixed by the code. I tested the fix with various text input fields in past 2 days. I haven't found problem yet. ( (In reply to comment #2) > (From update of attachment 150620 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=150620&action=review > > could you confirm the original problem fixed by the code you are changing does not regress? > > > Source/WebKit/blackberry/ChangeLog:6 > > + Reviewed by Antonio Gomes. > > I have not reviewed it :) > > Tools can fill it up for you
Comment on attachment 150631 [details] Patch It turns out the question Antonio asked still needs to be answered. Also the ChangeLog can be cleaned up a bit more, so far Antonio did not really review, so we should not mention him in the ChangeLog.
Hi Antonio, I must misunderstood "original problem" in my last comment. What is the original problem you talked about?
Created attachment 150652 [details] Patch
(In reply to comment #8) > Hi Antonio, > I must misunderstood "original problem" in my last comment. What is the original problem you talked about? I mean, without your change, do a git blame to these lines you are changing, and check what was the problem it was trying to fix, when we decided to not work with shadow content. Basically, I am wondering if changing it back to how it was regresses the original issue fixed by these lines.
Comment on attachment 150652 [details] Patch LGTM.
The Original PR is #127814. It just said that text fields from fat finger should all ASSERT(!node->isInShadowTree()). But specific reason had not been mentioned.
Invalid, the field should not be read only in the first place. Investigation is continuing, but this is not the desired approach.