Zack Weinberg wrote:- > I agree with you that (2) is a bad solution. However, I don't like > (1) either. It is a major quiet change to the semantics of Objective > C. I have no way of evaluating how widespread this sort of thing is, > so I'm going to be really conservative. > > Also, looking at cpplex.c and c-parse.in, I think going for (3) might > actually simplify the code. What do you think of the appended patch? > With it, the preprocessor crunches Nicola's test case correctly. > However, this only exposes another bug - see separate report. Two things: 1) Is it a good idea to introduce yet another token type, when CPP_OTHER would do (from what I can see)? 2) Don't you just drop stray @ tokens, in e.g. int x = y @ + z; without a diagnostic? (And what about in C?) Neil.
