Detecting parse warnings, parse errors, and build failures

Different incompatibilities can occur between differing dialects of C and especially C++, which result in parse errors and the cov-build command compiling less than all of the source code. You do not need all of the source code compiled to analyze the code for defects. However, the cov-analyze command analyzes only the files that cov-build was able to compile successfully.

The cov-build command, by default, is considered to be successful if it compiles 95% or more of the compilation units. You can change this percentage with the --parse-error-threshold option. For example, if you want cov-build to return without a warning only if 100% of the code compiles, add the following option to the cov-build command:

cov-build --dir <intermediate_directory> --parse-error-threshold 100

The more compilation units that you can compile without parse errors, the more code you can analyze. To improve the analysis, you can fix or work around many or all of the parse errors.

Sometimes the compiler can recover from a parse error. When the compiler recovers from an error, the compilation unit is compiled successfully, but the function that has the parse error cannot be analyzed. You can see these warnings (as RW.* checkers) in the Coverity Connect when you use the --enable-parse-warnings option to the cov-analyze command. To see cases when the compiler could not recover from errors, you should also specify the --enable PARSE_ERROR option to cov-analyze.

A variety of problems found by the Coverity compiler are called parse warnings, which you can see in the Coverity Connect (as PW.* checkers) when parse warnings are enabled. Parse warnings can show simple problems in the code, or can be signs of deeper defects. You can change which parse warnings are exposed as defects by creating a configuration file. A sample file is provided at <install_dir>/config/parse_warnings.conf.sample. For more information, see Coverity Command Reference.

If the compiler finds non-standard code, and it can infer what is intended by that code, the compiler generates a semantic warning, which you an see in the Coverity Connect (as SW.* checkers) when parse warnings are enabled.

The cov-build command returns a non-zero exit code when either there is a fatal error while attempting to initialize the cov-build state before launching the command, or when there is a non-zero exit code from the build command specified on the command line. In the case that there are build failures due to incompatibilities between the Coverity Analysis compiler and the source code being analyzed, if the error does not cause the native compiler to fail and the build to exit, cov-build will not exit with a non-zero status code. You can change this behavior by using the option --return-emit-failures.

For details about how to handle and resolve parsing incompatibilities, see Compiler-specific configurations.