This makes the translated output obey the HGENCODING environment
variable or the preferred encoding as set by the LANG or LC_ALL
environment variables.
Python 2.4 has a lgettext method which is similar, except that it
doesn't know about HGENCODING or the settings in .hgrc.
This default is to look for /usr/share/locale/xx/LC_MESSAGES/hg.mo for
language xx, but this code will instead do the lookup from locale/ or
mercurial/locale/ relative to the root of the Mercurial source tree.
The new demandimport eliminates most of the gettext performance hit of
importing gettext, which was doubling the run time for 'hg'.
Now, with 'hg tip':
dummy: .09530s no translations: .09638s translations: .09755s
