Bug#950972: press: Broken/mangled space characters in 10.3 and 9.12 point release announcements
Just noticed that in the release announcement for the 10.3 and
9.12 announcements, there seem to be broken spaces in the generated
table between the source package names and the reference markers.
Just checked quickly, in the script there is a U+00A0 (0xc2 0xa0) which seem to
cause the issue. If I replace the space with "normal" space U+0020, then the
issue disapear. The issue at least is triggerable as well with older issues not
only the recent 2020 ones.
The ‘use utf8’ declares that Perl source file's encoding is UTF-8.
The 0xa0 present in $link_format's definition hex-dumped above is
correct because it's U+00A0 encoded as 0xc2 0xa0. The source has other
non-ASCII characters correctly UTF-8 encoded, e.g. $openquote can
I think the problem lies with whatever takes this scripts output and
puts it into a MIME email message. This could be confirmed by running
the script in the same manner as for the email, but saving the output in
a file to be examined for UTF-8 correctness.
$ printf 'a\xa0b\n' | iconv -f utf-8 -t utf-8
aiconv: illegal input sequence at position 1