The English messages are compiled into man, and are used
when it cannot find a message catalog.
Before printing a message, man does a catopen("man_messages", 0);
which means that it tries to open the file obtained from the environment
variable NLSPATH by substituting "man_messages" for %N, and $LC_MESSAGES
for %L (older implementations use $LANG instead)
When NLSPATH is not set, the value
is used. If no information about the language is available, "C" is used.
Because the catalog routines are not generally available, I have enclosed
a copy of the gencat source.
If you add a new message catalog, say for language da, then verify that
the labels are used correctly by doing
../src/makemsg mess.en x.en.h x.en.c
../src/makemsg mess.da x.da.h x.da.c
diff x.en.h x.da.h
The files x.en.h and x.da.h should be identical.
Recent gencat wants to know what codeset the messages are in.
Please report incorrect codesets to firstname.lastname@example.org.