On Tue, Jul 30, 2013 at 02:00:09PM +0200, Ulrike Fischer wrote: > while with xetex one gets with both inputs U+00F6.
Note that what you seeing in XeTeX is not input normalization (XeTeX supports input normalization but it is not on by default), but is rather done at the text layout level by HarfBuzz which is a bit clever in its choice of normalization depending on what glyphs the font actually has, and it can even decompose composed characters if the font has the base and the combining mark but not the composed form for example, and it has some exceptions for some writing systems where certain decompositions give undesired results. In the PDF output it does not matter of course, but it should be different on macro level for example since input normalization is done when reading the input files. Regards, Khaled
