[CMake] bug ? changed behaviour of ADD_CUSTOM_COMMAND

[Anton Deguet]

It might be worth putting this in the SWIG section of the WiKi pages.
This is pretty useful to start a Python (or whatever script) which
relies on a SWIG wrapped library.

Anton


On Fri, 2005-09-16 at 09:22 -0400, Brad King wrote:
> Alexander Neundorf wrote:
> > Hi, 
> >  
> > since Makefile generator 2 the following cmake code doesn't work anymore: 
> >  
> > ----------8<-----8<--------8<------------- 
> >  
> > ADD_CUSTOM_TARGET( world echo "world" DEPENDS hello) 
> >  
> > ADD_CUSTOM_TARGET( hello echo "hello" ) 
> >  
> > ----------8<-----8<--------8<------------- 
> >  
> > Currently "make world" leads to  
> > make[3]: *** No rule to make target `hello', needed by `world'.  Stop. 
> >  
> > although the target "hello" is defined with the next command. 
> > With cmake 2.0 this did work and "make world" executed first "hello" and 
> > then "world". 
> > Is this change intentional ? 
> 
> The original behavior was not intentional.  The DEPENDS option of 
> ADD_CUSTOM_TARGET is meant for dependencies on the outputs of 
> ADD_CUSTOM_COMMAND.  It was just coincidence that the generated makefile 
> had the right dependencies, and I don't think it worked in other generators.
> 
> The way to add dependencies between top-level targets is with the 
> ADD_DEPENDENCIES command:
> 
> ADD_CUSTOM_TARGET( world echo "world" )
> ADD_CUSTOM_TARGET( hello echo "hello" )
> ADD_DEPENDENCIES( world hello )
> 
> -Brad
> _______________________________________________
> CMake mailing list
> CMake at cmake.org
> http://www.cmake.org/mailman/listinfo/cmake
> 
-- 
Anton Deguet <anton at cs.jhu.edu>
ERC CISST Johns Hopkins University