On Wed, 1 Dec 1999, Brian J. Fox wrote: > Is it possible that you are somehow using Cygwin in a mode where the > default character type is a wide 16-bit character (i.e., Unicode)? > > No, the right thing is to use O_BINARY in the call to read();. Are you saying that the default mode transparently expands characters into wide characters on Cygwin? The current assumption is that `stat' never underestimates the amount of bytes you wind up with after reading the file, it can only overestimate it. If that breaks in the case in point, then something should be done about it. But I'm not sure O_BINARY is the right thing in this case.
