[Libreoffice-commits] core.git: FastSaxSerializer: SAL_WARN() when writing invalid XML characters

[Stephan Bergmann]

On 03/01/2017 12:05 PM, Eike Rathke wrote:
> On Wednesday, 2017-03-01 10:34:04 +0100, Stephan Bergmann wrote:
>> (1)  If the input is assumed to be an arbitrary sequence of Unicode scalar
>> values (i.e., may contain noncharacters, even despite the caveat that those
>> should never be interchanged), the below invalidChar handling might want to
>> also watch out for U+FFFE and U+FFFF.
>
> Also if UTF-8 encoded? (as we write OString/chars there..)

Yes, the XML requirement is on the Unicode (or ISO/IEC 10646) 
characters, regardless how they're encoded in a given file.  Though it's 
probably a bit difficult to cram that check into FastSaxSerializer's 
design.  And, again, may even not be relevant if the input must not 
contain any noncharacters anyway.  (In configmgr, I didn't bother to 
ensure that at any higher abstraction level, and simply make sure 
arbitrary sequences of Unicode scalar values are properly encoded for 
XML's requirements.)