Integrating Coverity Analysis into the build environment—'cov-build'

The cov-build command integrates Coverity Analysis with a build system, usually without any modifications to the build system itself. Using cov-build is the preferred method of build integration. Figure 2 shows the basic process that cov-build uses to piggyback on a build system to produce the intermediate data. This intermediate data can then be analyzed to produce defect reports. For information about alternative build integration commands, see Alternative build command: 'cov-translate'.

After the cov-config.xml file is created, you can run the cov-build command by placing it in front of your usual build command. The required --dir option specifies the intermediate directory.

If the build command depends on features of the command shell that usually invoke it, such as certain shell variables or non-alphanumeric arguments, invoke the build command with a wrapper script. This method preserves the original behavior, since the build command is directly invoked by the type of shell on which it depends.

For example, if the normal invocation of a Windows build is:

> build.bat Release"C:\Release Build Path\"

use:

> cov-build --dir <intermediate_directory> wrapper.bat                  

where wrapper.bat is an executable command script that contains the original and unmodified build command.

On Windows systems, specify both the file name and extension for the build command when using cov-build.

For example:

> cov-build --dir <intermediate_directory> custombuild.cmd

Because cov-build uses the native Windows API to launch the build command, the appropriate interpreter must be specified with any script that is not directly executable by the operating system. For example, if the normal invocation of a build within Msys or Cygwin is:

> build.sh

prefix it with the name of the shell:

> cov-build --dir <intermediate_directory> sh build.sh

Similarly, if a Windows command file does not have Read and Execute permissions, invoke it as:

> cov-build --dir <intermediate_directory> cmd /c build.bat

The time that it takes to complete a build increases when you use cov-build because after the normal build runs, the Coverity compiler parses the same files again to produce the intermediate data. Consider the following factors that can increase build times with cov-build:

• The intermediate data directory is on a network mounted drive. Coverity Analysis creates many files and subdirectories in the intermediate directory, and these operations can be slow on network file systems. Using an intermediate directory on a local disk can eliminate this bottleneck. On Windows, you must use a local drive for the intermediate directory (Windows shared network drives are not supported for the intermediate directory).
• cov-emit does not take advantage of pre-compiled headers.

If the speed of cov-build is prohibitively slow when compared with your normal build time, one possible solution is to use more processes to parallelize the build. To see how to do so without altering your build scripts, see the section describing record/replay.