Hi Giuseppe (and anyone), wget's doc/Makefile.am assumes that --utf8 is supported:
$(MAN): wget.pod $(POD2MAN) --center [...] --utf8 $? > $@ (It's also assuming GNU make with the $? and $@, I believe, but never mind, that's a different issue.) Unfortunately, not all pod2man's have --utf8. It was introduced in a somewhat recent version of Perl. Exactly which version does not matter. Thus, the main bug: as far as I can see, the standard build process (configure && make install) should not invoke pod2man at all, any more than it should invoke makeinfo, and for all the same reasons. That is, the tarball should ship with the pod2man-generated file, just as it ships with the makeinfo-generated file. Alternatives: - configure could check if pod2man supports --utf8 and avoid using it if unsupported. - not that I expect you to change it, but the conversion of the Texinfo manual to POD seems baroque and unnecessary to me. You could have a minimal man page created with help2man instead, like most GNU programs. - I have even less expectation of you changing this, but the usage of @documentencoding UTF-8 (when unnecessary) in Texinfo is unfriendly to non-UTF-8 programmers. UTF-8 quotes appear as binary garbage in the C locale, thus the manual becomes unusable, since quotes are so pervasive in Texinfo output. (I'll spare you the reasons why I use LC_ALL=C, but will ask you to believe me, I don't do it lightly.) At any rate, I don't see that UTF-8 actually gains anything here (http://git.savannah.gnu.org/cgit/wget.git/commit/?id=8d4bb928b921b7620f1b3d61f04ed848138c3d7c). As far as I can see, the names can be represented perfectly well using Texinfo -- a 7-bit ASCII input file using Texinfo commands and no @documentencoding command should work fine. (If there are some diacritics unsupported by Texinfo, we can change that.) The HTML/PDF output will be correct and have the accents, and the plain text output will do as well as it can. What's unnecessary, indeed undesirable, are the @iftex conditionals that you removed. If the @iftex branches were restored over the direct UTF-8 input, all the problems would go away ... --best, karl.