Added

doc/lint/chap2

@@ -0,0 +1,57 @@
+.NH
+Outline of the program
+.PP
+The program can be divided into three parts. A first pass, which
+parses C program files and outputs definitions, a second pass which
+processes the definitions and a driver,
+which feeds the set of files to the first pass and
+directs its output to the second pass. Both passes produce the
+warnings on standard error output, which are redirected to standard
+output by the driver.
+.PP
+The first pass is based on an existing C front end, called
+.I cem
+[3].
+.I Cem
+is part of the Amsterdam Compiler Kit (ACK), as described in [4].
+.PP
+Most of the code of
+.I cem
+is left unchanged. This has several reasons. A lot of work, which
+is done by
+.I cem
+, must also be done by
+.I lint.
+E.g. the lexical analysis, the macro expansions,
+the parsing part and the semantical analysis.
+Only the code generation part is turned off.
+An advantage of this approach is, that a person who understands 
+.I cem
+will not have to spend to much time in understanding
+.I lint.
+.PP
+All changes and extensions to
+.I cem
+can be turned off by not defining the compiler directive
+.ft CW
+LINT.
+.R
+Compiling should then result in the original C compiler.
+.PP
+The second pass is a much less complex program.
+It reads simple definitions generated by the first pass and
+checks their consistency.
+This second pass gives warnings
+about wrong usage of function arguments, their results and
+about external variables, which are used and defined in more
+than one file.
+.PP
+The driver is a shell program, to be executed by the
+.UX
+shell
+.I sh.
+It executes the two passes and let them communicate through a
+filter (sort).
+Actually it is simplex communication: the first pass only talks to
+the second pass through the filter.
+.bp