On 12/30/11 12:21 AM, Eli Zaretskii wrote: >> Date: Thu, 29 Dec 2011 14:29:24 -0800 >> From: Daniel Colascione >> Cc: emacs-devel@gnu.org >> >> My initial revisions actually did exactly what you suggest, but I >> realized that the solution was more complex and didn't actually have >> any benefit. > > Can you describe the complexity in this solution? configure already knows what window system we're using, so it can set a macro to the name of the correct header without further logic. Nobody else has to be aware of the set of possible window systems. On the other hand, using a fixed "dispatch header" --- i.e., one that contains a bunch of preprocessor branches and include directives --- forces us to create yet another place that has to know about all possible window systems. >>> Also "TERM" does not look like a good prefix >>> in this case, it's meaning might be confused with the TERM environment >>> variable (nsterm/w32term/xterm are not that great either, but better not >>> propagate the confusion). >> >> "Term", I think, it pretty clear in the context of Emacs. Using a >> different name for the header constant wouldn't change the names of all >> the datatypes in that header. It's better to at least be consistently >> confusing. >> >> There's XTERM_HEADER, but this name has other issues. > > How about WINSYS_HEADER? That's fine.