Provided by: ftnchek_3.3.1-6_amd64
NAME
dcl2inc - postprocess ftnchek .dcl files to create separate INCLUDE files
SYNOPSIS
dcl2inc *.dcl
DESCRIPTION
dcl2inc postprocessing declaration files output by ftnchek(1), replacing unique COMMON block definitions by Fortran INCLUDE statements. For each input .dcl file, a modified output .dcn file is produced, together with include files named by the COMMON block name, with filename extension .inc. In addition, dcl2inc produces on stdout a list of Makefile dependencies for the UNIX make(1) utility. These can be appended to the project Makefile to ensure that any subsequent changes to .inc files provoke recompilation of source files that include them. dcl2inc warns about COMMONs which differ from their first occurrence, and simply copies them to the output .dcn file, instead of replacing them with an INCLUDE statement. Thus, any COMMON statements that are found in the output .dcn files should be examined carefully to determine why they differ: they may well be in error. Replication of identical data, and bugs arising from subsequent modification of only part of it, is a significant reason why Fortran programming projects should require that COMMON declarations occur in separate include files, so that there is only a single point of definition of any global object. Even though the Fortran INCLUDE statement was tragically omitted from the 1977 Standard, it has long been implemented by virtually all compiler vendors, and is part of the 1990 Standard. In practice, there is therefore no portability problem associated with use of INCLUDE statements, provided that one avoids nonportable file names. As long as the code obeys Fortran's limit of six-character alphanumeric names, the filenames generated by dcl2inc will be acceptable on all current popular operating systems. Fortran's default, or IMPLICIT, variable typing is deprecated in modern programming languages, because it encourages sloppy documentation, and worse, bugs due to misspelled variables, or variables that have been truncated because they extend past column 72. If all variables used are explicitly typed, and a compiler option is used to reject all program units with untyped variables, variable spelling and truncation errors can be eliminated. Variable declarations that have been produced automatically by a tool like ftnchek(1) or pfort(1) have a consistent format that facilitates application of stream editors (e.g. to change array dimensions or rename variables), and simple floating-point precision conversion tools like d2s(1), dtoq(1), dtos(1), qtod(1), s2d(1), and stod(1).
CAVEAT
The current version (2.9) of ftnchek(1) does not produce Fortran EQUIVALENCE statements in .dcl files, so you must be careful to preserve them when replacing original declarations with new ones from .dcl or .dcn files.
SEE ALSO
d2s(1), dtoq(1), dtos(1), ftnchek(1), make(1), pfort(1), qtod(1), s2d(1), stod(1).
AUTHOR
Nelson H. F. Beebe, Ph.D. Center for Scientific Computing Department of Mathematics University of Utah Salt Lake City, UT 84112 Tel: +1 801 581 5254 FAX: +1 801 581 4148 Email: <beebe@math.utah.edu>