Stephan Witt wrote: > Am 24.01.2016 um 13:00 schrieb Stephan Witt <st.w...@gmx.net>: > I’ve tried to follow the tokenizing and parsing program flow. > > While processing the „latin1“ option of „inputenc“ the encoding of > the idocstream is changed. (Line 395 of support/docstream.cpp is called) > > Perhaps this is the problem.
Good guess. What happens if you comment out the line with is.imbue() there? If it works with that line commented out, then we know that this is the problem. In that case it would not be a new bug at all. Georg
