I append a message from Robert Corbett of Sun Microsystems Inc.
([log in to unmask]) which he posted to Usenet earlier this week but
which did not seem to attract any reaction. It seems an excellent
question to pose, and it's nice to be asked for our opinions (but note the
disclaimer at the bottom). With his permission I am forwarding it to this
group where it may get a more considered response. Apologies to those who
have seen it before on comp.lang.fortran.
-Clive Page
=====================================================================
The 1.x releases of Sun f90 have placed information about modules
defined in a source file name /filename/./f90-suffix/, where
/filename/ and /f90-suffix/ are meta-names, in files named
"/filename/.M". Sun inherited this scheme from Cray Research,
on whose work Sun's compiler is based. During compilation, the
compiler searches all files that end with a .M suffix in the
directories implicitly or explicitly specified by the user
looking for information about modules.
It has not escaped my notice that most Fortran 90 compiler
vendors use a different scheme. The most popular scheme appears
to be to place information about a module named /module-name/ in
a file named "/module-name/.mod". Under this scheme, a single
source file can give rise to many .mod files.
To be frank, I prefer the .M scheme to the ,mod scheme. The .M
scheme allows make files to use a suffix rule to relate the module
information files to the corresponding source files. Under the
.mod scheme, no such correspondence exists. Also, the .M scheme
as it is implemented allows module information files to be
aggregated using concatenation. For example, if a library is
created from a set of Fortran 90 source files, a single .M file
for the entire library can be created simply by concatenating
the .M files created by the compiler for each of the source files.
Nonetheless, I recognize being out-of-step with the other Fortran 90
compiler vendors could create problems for our users. I would be
interested in knowing if people think Sun should retain the .M
scheme or if they think Sun should transition to the .mod scheme.
I am speaking only for myself, not for Sun.
Yours truly,
Robert Corbett
--
Clive Page, Internet: [log in to unmask]
Dept of Physics & Astronomy,
University of Leicester, Phone +44 116 252 3551
Leicester, LE1 7RH, U.K. Fax +44 116 252 3311
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|