Inference rules allow you to define rules that can be used for more than one file. When NMAKE32 encounters a target/dependency statement with no commands, it looks for an inference rule that specifies how to create the targets based on the defined .SUFFIXES extensions, the defined inference rules, and the target/dependent base file names.

In order for an inference rule to be applied, three conditions must be met:

• The extensions of the target file and the dependent file must be defined in the .SUFFIXES: target

• There must be a valid inference rule with correct path specification defined.

• The target and dependent must have the same base file name.

A valid inference rule, with local directory path specification, starts in the first column of the description file, and consists of the dependent extension, followed by the target extension and a colon. The command block starts on the following line, and must be preceded by at least one space.

   .extension1.extension2:       [command..]

Given a target/dependent statement, with the same base file name, and valid file extensions in the .SUFFIXES: target, NMAKE32 will search for a defined inference rule, and execute the rule if available. If the target and all dependents exist, and a rule cannot be inferred, the target is assumed to be "up to date".

Example:

  .c.obj:
     icc -c -Fo$@  $<

  hello.obj : hello.c

Similarly, given a target without a dependent or a command block, NMAKE32 will infer a dependent file from the targets base file name and all inference rules defined with the target's file extension. The inference rules are searched in the order that they are defined in the .SUFFIXES target.

Example:

  .SUFFIXES: .c .asm .obj .exe

  .obj.exe:
     icc -fe$@ $<

  .c.obj:
     icc -c -f$@ $<

  .asm.obj:
       masm $<;

  target.exe: header.h
  target.obj:

In the preceding example, NMAKE32 will search for an inference rule that builds a .exe file, it will locate the .exe.obj rule, using the target filename and the source extension, target.obj will be inferred as a dependent. It will then check to see if it can locate an up-to-date target.obj on disk, or can build one. When NMAKE32 locates target.obj as a target defined in the description file, it will try to build it. In order to build this file it will use the same inference rule process to infer the dependent target.c (searching extensions in the order given by .SUFFIXES). If target.c did not exist on disk or could be built from the description file, the next inference rule evaluated would be .asm.obj:, and the next inferred dependent would be target.asm.