I want to valgrind my entire application, not just the small, successful use cases. For example, I expect my app to exit with code 1 when the CLI flag syntax is wrong. However, it is difficult to check that flow in my app for leaks, as valgrind returns the wrong exit code in this case. I expected --error-exitcode=1 --leak-check=full to specifically exit non-zero from valgrind when leaks are found, not when my app exits non-zero.

My app does not allocate any memory when the CLI syntax is wrong, so valgrind should simply exit 0. However, in practice, valgrind forwards my app's exit value instead.

I realize that this kind of situation is difficult to manage: What should valgrind do when the application has no leaks, but is otherwise broken, such as exiting non-zero for a valid input; or exiting non-zero because the app does not have executable bits / is not found on $PATH?

Would be nice to have a flag to indicate that an application can be found, can run, has no leaks, and yet is expected to exit non-zero, and so valgrind itself should exit non-zero if and only if leaks are found. Does that make sense?