sax/source/tools/fastserializer.cxx | 11 +++++++++++ 1 file changed, 11 insertions(+)
New commits: commit 84b79376d7bc57a3967d7203a7d51466625588e8 Author: Stephan Bergmann <[email protected]> AuthorDate: Thu Feb 3 13:50:55 2022 +0100 Commit: Stephan Bergmann <[email protected]> CommitDate: Fri Feb 4 09:25:31 2022 +0100 Add a clarifying comment regarding the use of UTF-8 ...justifying that 2f3a0bfbfe110c0837b3c7e04f9ad0969d6e56e4 "tdf#147088: Also handle U+FFFE, U+FFFF invalid XML 1.0 characters" added code that assumes `string` is UTF-8 while carelessly removing the "assuming we're writing UTF-8" disclaimer comment that had been added with 8b25b67d5268abbb260da968cc23b6f6c8dd31af "escape invalid XML characters with _xHHHH_ when writing escaped" Change-Id: I0866da2bbbc536b2feb977c35b164459b745d918 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/129422 Tested-by: Jenkins Reviewed-by: Stephan Bergmann <[email protected]> diff --git a/sax/source/tools/fastserializer.cxx b/sax/source/tools/fastserializer.cxx index 8dcc308a836e..f4763dfa7c5d 100644 --- a/sax/source/tools/fastserializer.cxx +++ b/sax/source/tools/fastserializer.cxx @@ -106,6 +106,17 @@ namespace sax_fastparser { /** Characters not allowed in XML 1.0 XML 1.1 would exclude only U+0000 + + This assumes that `string` is UTF-8, but which appears to generally be the case: The only + user of this FastSaxSerializer code is FastSerializerHelper, and when its constructor + (sax/source/tools/fshelper.cxx) is called with bWriteHeader being true, it calls + FastSaxSerializer::startDocument, which writes sXmlHeader claiming encoding="UTF-8". The + only place that appears to construct FastSerializerHelper appears to be + XmlFilterBase::openFragmentStreamWithSerializer (oox/source/core/xmlfilterbase.cxx), and it + only passes false for bWriteHeader when the given rMediaType contains "vml" but not "+xml" + (see <https://git.libreoffice.org/core/+/6a11add2c4ea975356cfb7bab02301788c79c904%5E!/> + "XLSX VML Export fixes", stating "Don't write xml headers for vml files"). But lets assume + that even such Vector Markup Language files are written as UTF-8. */ template<typename Int> static std::optional<std::pair<unsigned, Int>> invalidChar( char const * string, Int length, Int index )
