Description of problem: How to convert a Czech file full of diacritic signs into ASCII without mr. Google? It is impossible without knowledge of the undocumented switch. $ cat utf8.txt áäâăąćčçďđéěëęíîĺľłńňóôöőŕřśšşťţúůüűýźžż ÁÄÂĂĄĆČÇĎĐÉĚËĘÍÎĹĽŁŃŇÓÔÖŐŔŘŚŠŞŤŢÚŮÜŰÝŹŽŻ $ iconv -f UTF8 -t ASCII utf8.txt iconv: illegal input sequence at position 0 $ iconv -f UTF8 -t ASCII//translit utf8.txt aaaaacccd?eeeeiill?nnoooorrsssttuuuuyzzz AAAAACCCD?EEEEIILL?NNOOOORRSSSTTUUUUYZZZ Version-Release number of selected component (if applicable): glibc-common-2.5-10.fc6 How reproducible: always
man-pages aren't part of glibc and iconv -l is really supposed to give you list of the charsets, not supplement a manual page and tell you about all possible modifiers. //TRANSLIT and //IGNORE are documented at: http://www.gnu.org/software/libiconv/documentation/libiconv/iconv_open.3.html But it can be a name of a user provided shared library that implements special translation as well.
Yeah, glibc is completelly different thing :-) I simply issued the following command: $ rpm -qf /usr/bin/iconv glibc-common-2.5-10.fc6 Maybe `iconv --list` should give a hint, because I can't imagine to integrate a such marginal thing into "iconv" man page.
The fix as an example to the man page seems to be good and painless solution ... My proposal: --- man 2007-07-09 19:29:29.000000000 +0200 +++ fix4man 2007-07-09 19:31:35.000000000 +0200 @@ -30,6 +30,13 @@ Print progress information. +EXAMPLES + Following examples convert from code set ISO88592 "input.txt" to UTF8 code set or ASCII and stores the result as "output.txt". + + iconv -f ISO88592 -t UTF8 < input.txt > output.txt + + iconv -f ISO88592 -t ASCII//TRANSLIT < input.txt > output.txt + AUTHOR iconv is written by Ulrich Drepper as part of the GNU C Library.
Iconv man page provided by man-pages doesn't mention //TRANSLIT, //IGNORE but also -c and -s. If you want I could provide new iconv man page heavily based on the one from libiconv which will match latest iconv from glibc, are you interested?
Thanks Josef, the patch is applied in the last version of man-pages in fedora (man-pages-2.63-3.fc8). Thanks very much Marcin, it would be great if you contact Michael Kerrisk - he is the man-pages upstream maintainer and he will be the best person who know the recent status of man pages and who could use your work in the upstream version. If there is any problem with contacting Michael Kerrisk, write here a comment or write me an e-mail. Thanks again.
The problem is that iconv(1) I'm telling about, and which Josef's patch has been included came from man-pages-extralocale.tar.bz2. iconv(1) man page in this archive are from Debian's glibc package (BTW Debian glibc have newer version of this man page). iconv man page from man-pages is iconv(1p) (note the 'p'), but when user type "man iconv" iconv(1) is selected. iconv(1p) isn't synced with glibc's iconv either. I was about to contact man-pages upstream maintainer, but after checking this I wrote in this bug about new iconv(1) matching latest iconv from glibc :)