isofs: use unsigned char types consistently
Based on the discussion about the signed character field for the year, I went through all fields in the iso9660 and rockridge standards to see whether they should used signed or unsigned characters. Only a single 8-bit value is defined as signed per 'section 7.1.2': the timezone offset in a timestamp, this has always been handled correctly through explicit sign-extension. All others are either '7.1.1 8-bit unsigned numerical values' or composite fields. I also read the linux source code and came to the same conclusion, also I could not find any other part of the implementation that actually behaves differently for signed or unsigned values. Since it is still ambigous to use plain 'char' in interface definitions, I'm changing all fields representing numbers and reserved bytes to the unambiguous '__u8'. Fields that hold actual strings are left as 'char' arrays. I built the code with '-Wpointer-sign -Wsign-compare' to see if anything got left out, but couldn't find anything wrong with the remaining warnings. This patch should not change runtime behavior and does not need to be backported. Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jan Kara <jack@suse.cz>
Showing
Please register or sign in to comment