Frank A Uepping <[EMAIL PROTECTED]> writes: > Should the developer analyze its software for all the required > features it needs and fit out configure.ac with a test for it (which > will make configure fail if the test fails).
> Or should the developer distribute its software with a minimal > configure.ac and let the compilation fail if something is incompatible. > What is the right strategy? I generally test for in configure and fail for things like external libraries that are fairly likely to be missing if someone didn't follow instructions, since otherwise the error messages are too confusing. But for actual porting problems (different headers, missing functions, etc.), I just let the compilation fail. -- Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/>
