Fedora recently added this feature. I was able to hack it into the lesspipe.sh that is part of the ebuild, but had to prevent the early exit. No time to figure out how to integrate it with that at the moment. --- /usr/portage/sys-apps/less/files/lesspipe.sh 2011-01-19 21:31:13.000000000 -0600 +++ /home/ssb/bin/./lesspipe.sh 2011-04-28 18:07:18.029585226 -0500 @@ -91,7 +91,7 @@ *.ps|*.pdf) ps2ascii "$1" || pstotext "$1" || pdftotext "$1" ;; *.doc) antiword "$1" || catdoc "$1" ;; *.rtf) unrtf --nopict --text "$1" ;; - *.conf|*.txt|*.log) ;; # force less to work on these directly #150256 +# *.conf|*.txt|*.log) ;; # force less to work on these directly #150256 ### URLs ### ftp://*|http://*|*.htm|*.html) @@ -195,6 +195,17 @@ # Maybe we didn't match because the file is named weird ... 1) lesspipe_file "$1" ;; esac + if [ -x /usr/bin/file -a -x /usr/bin/iconv -a -x /usr/bin/cut ]; then + case `file -b "$1"` in + *UTF-16*) conv='UTF-16' ;; + *UTF-32*) conv='UTF-32' ;; + esac + env=`echo $LANG | cut -d. -f2` + if [ -n "$conv" -a -n "$env" -a "$conv" != "$env" ]; then + iconv -f $conv -t $env "$1" + exit $? + fi + fi # So no matches from above ... finally fall back to an external # coloring package. No matching here so we don't have to worry Reproducible: Always Steps to Reproduce: 1.any arbitrary file with UTF-16 content is displayed without decoding 2. 3.
post patches as attachments. bugzilla comments corrupts them. hardcodes paths are never acceptable. the invocation is also wrong as it will fail on files that begin with a dash. the logic also breaks as the comment indicates. the LANG parsing is wrong. pretty sure it should be LC_CTYPE. further, splitting on "." doesnt work for all possible names. it also will wrongly exit if the locale is invalid, or there is a conversion error.
This bug has been open, awaiting a response for over two years now. If you are still interested in having this feature added, please attach a patch that is modified in response to vapier's feedback. (I am closing out old base-system bugs to get the list to a manageable size.)
Created attachment 361536 [details, diff] cleaned up patch Cleaned up patch based on your comments. Sorry I didn't ever get back to you. I think LANG needs to be included since that's the last resort, but from what I've read it should try to use in order of preference: LC_ALL LC_CTYPE LANG. I'm not sure what to do about the parsing problem with using ".", but it seems reasonable to me for it to fall through if it can't find a second field after a ".". I also changed it to fall through if iconv returns anything other than success since there are some conversions that it doesn't like such as ca_ES.UTF-8@valencia. I wasn't sure what to do about the error message in view of falling through - maybe it should go to /dev/null.
can you post some sample files and the locales you're using to view them ?
