The Calendar library from http://www.lri.fr/~signoles/prog/calendar/doc/
contains a module (Period) which has an interface (period.mli) but no
implementation.

There's apparently nothing wrong with that, but the output of
tools/objinfo for calendar.cma shows only the list of modules which
have implementations.  For example:

  Unit name: Time
  Interfaces imported:
        71f888453b0f26895819460a72f07493        Pervasives
        1876ce678cfa5a1961fac21850f71efd        Unix
        1b657b86bb0e7ba36acf02910897b2eb        Time
        8d84727c6f398e1a8b710d8161e26479        Time_Zone
        d99e051c4c63bb80cd2586ea99584429        Period
  Uses unsafe features: no

"Period" is not shown as a separate module (ie. there is no "Unit
name: Period") in the output.

The problem for me is I'd like to determine which modules are provided
by a library, and which modules a library depends upon.  I can't tell
that from objinfo because of modules like Period.

I had a look at the implementation, tools/objinfo.ml and the
definition of compilation_unit and library in bytecomp/emitcode.mli,
and I can't really see how even the compiler knows about these types
of modules.  Why doesn't it try and fail to link to Period when
loading calendar.cma?

Rich.

PS. Background here:
https://www.redhat.com/archives/fedora-devel-list/2007-May/thread.html#01234

-- 
Richard Jones
Red Hat