[CMake] Compilation failure req'd for successful test

[Brad King]

Walter E Brown wrote:
> As part of a project's testing, we need to ensure that a compilation 
> failure will correctly occur under specific circumstances.  In 
> particular, we wish at ctest time:
> 
>   - to compile (but not link) a specific program that we expect to fail 
> compilation, then
> 
>   - to capture the detailed output of that compilation into a file, and 
> then
> 
>   - to run some other program (perhaps a tool like grep or perhaps a 
> custom script) that inspects that attempted compilation's output for 
> certain characteristics of interest, and finally
> 
>   - to indicate the success or failure of the test based on the results 
> of such inspection.
> 
> Alas, it is unclear to us how to set this up; neither the book nor the 
> online help seem on point.  Can someone advise us, preferably with 
> sample cmake/ctest code that we can use as a model?

Currently there is no easy way to do this.  You may have to create a 
program that drives this process itself and then add the program as the 
test.  It can use "ctest --build-and-test ..." to drive the build and 
then capture the exit code and output.  You can look at 
CMake/Source/kwsys (and even copy the library) to get some nice 
cross-platform process execution code to help achieve this.

In the long run this should be a ctest feature accessible through an 
option like "--build-fail-compile".  Please add it as a feature request 
here:

http://www.cmake.org/Bug

Thanks,
-Brad