> It would be nice to have a comment in the patch file explaining why > you changed NAN to 0, and whether you sent the fix upstream, their bug > tracking id, etc. But not necessary. Forgot to mention: This is work in progress. The problem is simple, but the solution is not. Most likely this is a bug in NetBSD: There are (for that C-file) global variables initialized statically with the symbol "NAN". "NAN" sugests that this symbol is somthing static, a literal like "0.0" or "0xff". But in NetBSD NAN is a preprocessor macro that finaly resolves to the variable "__nanf" inside libc. So in NetBSD it is not possible to statically initialize a variable with NAN. Too me this looks more like a problem of NetBSD, not of the application. I am still not sure how to solve this. The real question is what C99 says about the use of NAN and if it's required to be a constant - that will determine if viking or NetBSD is wrong :-) Section 7.12 part 5 says that NAN is a macro expanding to a constant of type float that is a quiet NaN. So the original code is legit and NetBSD is buggy, I think.
Description: PGP signature