> >> I am not sure the following is necessary if the source files are built > >> in the same directory where they are needed, but for the case where > >> source files are built in a different directory (the case for my > >> generated header), you must have your executable or library target > >> depend on an empty custom target which in turn depends on your custom > >> command to insure the source files get built before they need to be > >> compiled.
I am doing that. Otherwise it will not be included in the dependency graph. > >> In any case, I do not think multiple invocations of cmake will be > >> necessary in order to use generated source files. > > > > But if I understand correctly then at cmake time the files which will > > have to be compiled are not yet known, right ? > > I think this is a problem. I know their names, since it is just a case of incrementing the number in the file's name. The problem is that, should I overcalculate N, make ought to complain about a file not found. If I undercalculate it, make will be skipping needed files. The simplified process is something like this: a.txt -> a.pl b.txt -> b.pl ... Processing each of these files: ?.pl -> big_file.isco -> big_file_1.isco, big_file_2.isco, ..., big_file_N.isco Then, for each of these: big_file_i.isco -> big_file_i.pl -> big_file_i.o Then the .o files follow the same path as the rest of the "normal" code. (these are prolog files, btw) > However, what then is the purpose of the SET_SOURCE_FILES_PROPERTIES > GENERATED property? I have never used that flag, but I cannot think of any > use for it except the present situation. I'm new to CMake, but I never set this, and all works fine in my generated files. Thank you. -- Gustavo _______________________________________________ CMake mailing list CMake@cmake.org http://www.cmake.org/mailman/listinfo/cmake