Eric Blake wrote: > Does this look correct to apply? The MKDIR_P changes are ok but are not yet necessary, as there is only 1 module that creates the 'arpa' subdirectory and only 1 module that creates the 'netinet' subdirectory.
The Makefile dependency would be highly annoying. Many users don't use the automake option 'no-dependencies', therefore use full dependencies. A rebuild or modification of a generated header file like string.h or unistd.h causes large parts of the package to be rebuilt. The user then sits there for minutes and minutes, waiting for gcc to do its job. For what benefit? Do you expect the generated header files to be different after the user added "-g" or removed "-O2" from the CFLAGS in his Makefile? No, the only users who do things that change string.h are us few, and we can utilize 0.5 seconds of our brain in order to determine which file to remove before doing make. Don't let the users suffer. > > adding a dependency on config.status doesn't seem to be the right thing to > > do. Yes, for the same reason. Bruno
