The file that contains the set of customized rules for NMAKE32 is a special file referred to as a description file. The default description file that NMAKE32 will search for is named makefile. The /f option is used to provide a description file name when the user does not want to use the default description file. NMAKE32 will return an error if a description file is not provided. A description file consists of the following elements:

• Description blocks
  
• Macro definitions
  
• Inference rules
  
• Directives

The following syntax rules apply to description files:

• A target/dependency statement must begin in column one.

• A command line must begin with a space; it cannot start in column 1.

• A backslash(\) at the end of a line tells NMAKE32 that the next line is a continuation of the current line.

• An unescaped pound sign (#) anywhere on a line in a description file indicates that the remainder of the line is a comment. Since the lines in an in-line file are processed by the command shell, specifying comments is specific to the shell in use. In OS/2, CMD.EXE uses rem.

NMAKE32 uses the target/dependency rules in the description file to determine if a target needs to be updated. All dependents are recursively checked for being out of date with respect to their dependents. Missing targets are considered to be out of date. If a target is determined to be out of date, the command associated with the target/dependency statement is executed. See the following example:

    program.exe:    program.obj  abcd.obj        ilink program abcd;

    program.obj:    program.c   xxx.h
        icc program.c

    abcd.obj:       abcd.c      xxx.h
        icc abcd.c

The above example description file says that program.exe depends on two files, program.obj and abcd.obj, and that they in turn depend on their corresponding files, program.c, abcd.c and a common include file, xxx.h.

When NMAKE32 is executed with the above description file, it would determine that program.exe was the main target and that abcd.obj and program.obj are the files that it depends on. NMAKE32 would then recursively determine that program.obj depends on program.c and xxx.h, and that abcd.obj depends on abcd.c and xxx.h. Thus, NMAKE32 constructs a graph of file dependencies. Starting from each node of the graph, NMAKE32 compares the modification date of each node (file) to its immediate descendants. If a file is older than one or more of its descendants, then the command associated with the file is executed to bring the file up to date.

When the above description file is processed by NMAKE32, assuming that each .obj file is older than the corresponding .c file, the following commands would be executed:

   icc -c program.c
   icc -c abcd.c
   ilink program abcd;

These are all the necessary commands to make program.exe from program.c, abcd.c, and xxx.h. If any of these commands encountered an error, NMAKE32 would immediately terminate and report which command failed.

If the file abcd.c was modified, then executing NMAKE32 again would only execute the commands necessary to bring program.exe up to date. The commands executed now would be:

   icc -c abcd.c
   ilink program abcd;

If the file program.exe was missing, executing NMAKE32 would execute the following command:

   ilink program abcd;

If the file xxx.h was modified, executing NMAKE32 would cause all the commands to be issued again because abcd.obj and program.obj depend on it.