1161807 - [RTL] changing textContent on dir=auto element from RTL to phone number does not update direction

Status: RESOLVED DUPLICATE of bug 1103348

(Core :: Layout, defect)

Description

[Michael Henretty [:mikehenrty][:mhenretty]]

Attachment: [HTML page] test case

When an element has dir=auto set, and is populated with RTL text (like Arabic), if you then set the textContent to a phone number (like [phone]), the text direction will stay in RTL mode. However, if you do not have Arabic text populated first and set the text to the phone number, the element will correctly use the LTR direction. See attached test case.

Comment 1

[Michael Henretty [:mikehenrty][:mhenretty]]

This is blocking a blocker. David, do you know what is going on here? It looks like we can work around this issue by temporarily setting an LRT textContent ("A" for instance) to the element before setting the phone number, but this would be less than ideal.

Comment 2

[Simon Montagu :smontagu]

The current behaviour in the testcase is strange and inconsistent and need to be fixed, but I'm not sure that the behaviour you want is what the spec says should happen: a phone number has no strong directional characters, so a <div> with dir="auto" whose text content is only a phone number has the same directionality as its parent. An <input> or a <textarea>, on the other hand, defaults to ltr if it contains no strong directional characters. 

https://html.spec.whatwg.org/multipage/dom.html#the-dir-attribute

Ehsan, do you remember why the spec makes this distinction?

Comment 3

[Jonathan Kew [:jfkthame]]

FWIW, the behavior here sounds like it might be related to the case described in bug 1103348.

Comment 4

[Simon Montagu :smontagu]

Wow, yes. I had no memory of that bug at all, in spite of having written a lengthy comment in it. Apparently international travel causes retrogressive amnesia for me :(

Comment 5

[Julien Wajsberg [:julienw]]

(In reply to Simon Montagu :smontagu from comment #2)
> The current behaviour in the testcase is strange and inconsistent and need
> to be fixed, but I'm not sure that the behaviour you want is what the spec
> says should happen: a phone number has no strong directional characters, so
> a <div> with dir="auto" whose text content is only a phone number has the
> same directionality as its parent.

But that's not how gecko works right now, see http://jsbin.com/jebahefunu/1/edit?html,output for example. When we use "dir=auto" the phone number is LTR in a RTL document.

This helps us a lot currently :) if the behavior changes we'll have to do a lot of fixes in the comms apps like SMS.

Comment 6

[(no longer active)]

(In reply to Simon Montagu :smontagu from comment #2)
> Ehsan, do you remember why the spec makes this distinction?

I kind of remember that there was a (good?) reason, but unfortunately don't remember anything more.

(In reply to Julien Wajsberg [:julienw] (PTO May 8th -> May 17th) from comment #5)
> (In reply to Simon Montagu :smontagu from comment #2)
> > The current behaviour in the testcase is strange and inconsistent and need
> > to be fixed, but I'm not sure that the behaviour you want is what the spec
> > says should happen: a phone number has no strong directional characters, so
> > a <div> with dir="auto" whose text content is only a phone number has the
> > same directionality as its parent.
> 
> But that's not how gecko works right now, see
> http://jsbin.com/jebahefunu/1/edit?html,output for example. When we use
> "dir=auto" the phone number is LTR in a RTL document.
> 
> This helps us a lot currently :) if the behavior changes we'll have to do a
> lot of fixes in the comms apps like SMS.

If I'm reading the spec wrong, this seems like a bug.  But I'm not sure why Chrome and Safari both behave the same.  :/

Comment 7

[(no longer active)]

This is a dupe of bug 1103348.  My patch for that bug fixes this too.