Static Analysis

What is Static Analysis?
Won't my compiler do syntax checking?
Does C++ make lint obsolete?
How do I separate the wheat from the chaff?
What's the difference between PC-lint and FlexeLint?

What is Static Analysis?

PC-lint and FlexeLint will look across your entire program that may consist of many C and/or C++ modules and find bugs, glitches, inconsistencies, and redundancies. It can find problems that may not be caught in initial testing. Some bugs lie dormant for months and years and their ultimate eradication can be time consuming and expensive. It pays to expose errata while the code is still fresh in the mind of the programmer and while the original programmer is still employed by the company responsible for the software.

Compilers do an excellent job of finding syntax errors and most will produce warnings. But PC-lint and FlexeLint begin where compilers leave off. PC-lint/FlexeLint contains numerous features such as precision tracking, initialization checking, value tracking, strong type checking and macro analysis that compilers do not have. Also PC-lint/FlexeLint looks across a set of modules to find intermodule inconsistencies and redundancies. This compilers do not do.

ANSI C and C++ have solved the problem for which Unix lint was originally developed, i.e., to ensure consistency of function calls across module boundaries. But most other sources of errors in C code remain in C++. These include uninitialized variables, order of evaluation dependencies, loss of precision, potential uses of the null pointer, etc. (See Representative Checks on the next page). In addition to problems inherited from C, C++ has added new opportunities and requirements for static checking. For example, it is now possible to report that a constructor has not initialized a member. This deficiency could not be reported for C code because responsibility for initialization is not uniquely assigned to a given function (i.e. there are no constructors in C). There have been numerous articles and books on the C++ language written by such prominent authors as Cargill, Koenig, Murray and Myers describing numerous do’s and don’ts. Although it is a formidable task for even the expert to hand check each line of source code for conformity to accepted principles of good design, such checks can be automated and most have been incorporated in PC-lint/FlexeLint.

We have taken several steps so that our users are not inundated by a mass of insignificant messages in which they may have no particular interest. All messages are carefully crafted to make use of all available information so that they have an inherently high wheat-to-chaff ratio. We have four levels of message (Error, Warning, Informational and Elective Note) and distinguish between two kinds of header (library headers vs. project headers). Messages can be suppressed by number, by symbol, by macro, by line and by header type. Through a selection of options, you in effect tailor your own subset of the C/C++ language.

PC-lint for C/C++  runs on Windows and MS-DOS.

FlexeLint is functionally identical but is available for a wider range of platforms (all flavors of Unix and Linux, VMS, OS-9, etc) and is distributed in "shrouded source" form.