[Development] Can we pass QTextDocument arbitary data?

[Lars Knoll]

> On 22 Jan 2026, at 02:25, Thiago Macieira <thiago.macieira at intel.com> wrote:
> 
> On Wednesday, 21 January 2026 15:17:50 Pacific Standard Time Albert Astals Cid 
> wrote:
>> We have found a garbage string that when passed to QTextDocument makes it
>> assert
> 
> Code in question is this:
> 
>    const bool isSurrogate = c.isHighSurrogate() && i < length - 1;
>    const char32_t ucs4 = isSurrogate
>                            ? QChar::surrogateToUcs4(c, string[++i])
>                            : c.unicode();
>    const QUnicodeTables::Properties *p = QUnicodeTables::properties(ucs4);
> 
> It doesn't verify that the code unit after the high surrogate is a low 
> surrogate. That must be why it ended up with a UTF-32 code unit outside of the 
> Unicode range.
> 
> This could be fixed in QTextEngine. But I would argue that passing corrupted 
> UTF-16 to any Qt API outside of QString itself is the bug.

I'd argue that Qt should not crash on invalid UTF16 in any case. There are only a handful of places where we use QChar::surrogateToUcs4() in Qt, those should be made safe IMO. Or fix surrogateToUcs4() to return the object replacement character in case the values are out of range.

Cheers,
Lars