/usr/share/postgresql_autodoc/html.tmpl specifies the HTML encoding
twice: once, near the top, as UTF-8 (correct) and once, near the
bottom of the <head/>, as latin1 (incorrect).

Please remove the second instance and ensure postgresql_autodoc
always produces correct UTF-8.

This has just now started to produce bogus input, namely when
nōn-ASCII chars are used in the database. For just umlauts:


~$ sudo su - postgres
$ createdb -E UTF-8 -O postgres -T template0 -l de_DE.UTF-8 tstdb
$ export LC_ALL=C.UTF-8
$ psql tstdb
psql (10.0)
Type "help" for help.

tstdb=# CREATE FUNCTION tst() RETURNS trigger LANGUAGE plpgsql AS $$ BEGIN
tstdb$#   RAISE EXCEPTION 'mäh';
tstdb$# END; $$;
CREATE FUNCTION
tstdb=# \q
$ postgresql_autodoc -d tstdb -f foo -t html
Producing foo.html from /usr/share/postgresql_autodoc/html.tmpl
$ fgrep -a -e \<meta -e RAISE foo.html
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
    <meta HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
  RAISE EXCEPTION 'm�h';
$ fgrep -a -e \<meta -e RAISE /usr/share/postgresql_autodoc/html.tmpl
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
    <meta HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">


When using Unicode characters beyond latin1, things get inconsistent:


$ psql tstdb
psql (10.0)
Type "help" for help.

tstdb=# CREATE FUNCTION tst2() RETURNS trigger LANGUAGE plpgsql AS $$ BEGIN
  RAISE EXCEPTION 'm€h';
END; $$;
CREATE FUNCTION
tstdb=# \q
$ postgresql_autodoc -d tstdb -f foo -t html
Producing foo.html from /usr/share/postgresql_autodoc/html.tmpl
Wide character in print at /usr/bin/postgresql_autodoc line 1616.
$ fgrep -a -e \<meta -e RAISE foo.html
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
    <meta HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
  RAISE EXCEPTION 'mäh';
  RAISE EXCEPTION 'm€h';