The judgment of "be:confirmed" is too early.

[bxi07354@…]

src/kits/interface/BTextView/TextView.cpp

BTextView::HandleInputMethodChanged(BMessage *message)

After InsertText() is done, it is necessary to judge "be:confirmed".

[jackburton]

More details please.

[bxi07354@…]

B_INPUT_METHOD_EVENT HandleInputMethodChanged

---

CURRENT

1. be:string="a" be:confirmed="false" InsertText("a");
2. be:string="ab" be:confirmed="false" DeleteText("a"; InsertText("ab");
3. be:string="abc" be:confirmed="true" return; OPINION
4. be:string="a" be:confirmed="false" InsertText("a");
5. be:string="ab" be:confirmed="false" DeleteText("a"; InsertText("ab");
6. be:string="abc" be:confirmed="true" DeleteText("ab"; InsertText("abc")

[jackburton]

Sorry, I meant: Why do you think we should change how it's done ? Does this cause any problem at all? IIRC me and Andrew (Bachmann) came to the current solution after careful testing with a couple input add-ons.

[jackburton]

1. be:string="a" be:confirmed="false" InsertText("a");
2. be:string="ab" be:confirmed="false" DeleteText("a"; InsertText("ab");
3. be:string="abc" be:confirmed="true" return;

Which IME is causing this ? With the ones we tested, the sequence is:

1. be:string="a" be:confirmed="false" InsertText("a");
2. be:string="ab" be:confirmed="false" DeleteText("a"; InsertText("ab");
3. be:string="abc" be:confirmed="false" DeleteText("ab"; InsertText("abc");
4. be:string="abc" be:confirmed="true" return;

[bxi07354@…]

Sorry. I cannot explain in English. example1. be:string="a" be:confirmed="false" InsertText("a"); 2. be:string="bc" be:confirmed="false" DeleteText("a"); InsertText("bc"); 3. be:string="edf" be:confirmed="false" DeleteText("bc"); InsertText("def"); 4. be:string="" be:confirmed="true" DeleteText("def");be:confirmed="true" shows present be:string. It is not previous one be:string. bool confirmed = false; if (message->FindBool("be:confirmed", &confirmed) == B_OK && confirmed) { fInline->SetActive(false); Refresh(0, fInline->Offset() + fInline->Length(), true, false); return; }It is moved to the 4470th line.

[jackburton]

What I'm trying to explain is that this implementation works fine in BeOS R5, tested with the R5 IME and with another IME which I can't remember. If isn't working correctly in Haiku, maybe it's an issue of our input server or the IME which you're testing with (I suppose it's Canna).

[bxi07354@…]

This problem is not in canna. It becomes a problem by skkIM. (​http://bebits.com/app/675)

[jackburton]

I'll try to have a look.

[korli]

Could you provide a patch ("diff -u") ? :)

[jackburton]

Jerome, please don't apply any patch yet. As you can see in the svn history, in rev 9761 we were handling the confirmation later, but we changed it in rev 9759 because of problems with some IME.

[bxi07354@…]

TextView.cpp

[jackburton]

Fixed by Axel in svn.