Version-Release number of selected component (if applicable): man-pages-1.53-3 (RedHat 9, fully up2date system) How reproducible: Always Description of problem: The following files: //generated via rpm -q --filesbypkg man-pages | sed "s@.*/usr@/usr@" | while read i; do zcat "$i" | iconv -f UTF-8 -t ISO-8859-1 2>/dev/null >/dev/null || echo "$i"; done /usr/share/man/man2/close.2.gz /usr/share/man/man2/getdomainname.2.gz /usr/share/man/man2/getrlimit.2.gz /usr/share/man/man2/madvise.2.gz /usr/share/man/man2/sysinfo.2.gz /usr/share/man/man2/time.2.gz /usr/share/man/man2/umask.2.gz /usr/share/man/man3/encrypt.3.gz /usr/share/man/man3/fclose.3.gz /usr/share/man/man3/fcloseall.3.gz /usr/share/man/man3/fflush.3.gz /usr/share/man/man3/lockf.3.gz /usr/share/man/man3/printf.3.gz /usr/share/man/man3/rand.3.gz /usr/share/man/man3/strtok.3.gz /usr/share/man/man3/toupper.3.gz /usr/share/man/man3/updwtmp.3.gz /usr/share/man/man4/st.4.gz /usr/share/man/man5/environ.5.gz /usr/share/man/man5/utmp.5.gz /usr/share/man/man7/glob.7.gz /usr/share/man/man7/hier.7.gz /usr/share/man/man7/iso_8859-1.7.gz /usr/share/man/man7/iso_8859-15.7.gz /usr/share/man/man7/iso_8859-2.7.gz /usr/share/man/man7/iso_8859-7.7.gz /usr/share/man/man7/koi8-r.7.gz /usr/share/man/man7/suffixes.7.gz fail conversion from UTF-8 to <anything> since they are not UTF-8 encoded. This is important since this step is performed by man resulting in "iconv: illegal input sequence at position ####" error messages from i.e. "man 2 close" et al. I've also seen this error in Polish language manual pages. Furthermore the above list may be incomplete as it only catches manpages with invalid chars (thus obviously not correct) and not possibly correct UTF-8 man pages which aren't encoded as UTF-8. Either /usr/bin/nroff should be changed to not use iconv -f UTF-8 or all man-pages should be converted to UTF-8 (or some auto-detection code?)... This is very annoying as it makes many man pages useless (LC_ALL et all settings change nothing as the problem lies in the input encoding, which isn't UTF-8 as expected and not the output encoding which can be changed via locale settings)
Still present in Fedora Core 1
To reproduce the bug, one can run "man iso_8859_1" in a gnome-terminal. The character table is just full of question marks instead of different characters.
The way encoding is handled has changed from RHL 9, which expected the source character set for the man pages from Western European language localed to be in ISO-8859-1 (which was then converted to UTF-8). They do need to be re-encoded.