> From: Diane Holt [mailto:[EMAIL PROTECTED]] I never meant for this to drag out for so long, but let me try one more increment. Maybe we can identity where our understandings of the meaning of "relative" diverge. (It might be in Windows vs. in Ant.) Isn't whether a pathname is relative defined by whether its meaning depends on something else (whether it is evaluated relative to something else)? If a pathname always designates the same file regardless of your current directory setting, it's absolute, right? If it depends on the current directory setting, even if only on the drive-letter portion of the current directory setting, it's not absolute, right?. If a pathname is not absolute, then it's relative, right? For example, the Windows pathname "\x" does not always designate the same file. It designates "C:\x" if the drive portion of the current directory setting is the "C:" drive, and it designates "D:\x" if the drive portion is "D:". Therefore, "\x" is not an absolute pathname, right? Therefore, it is a relative pathname, right? (Note that "relative" doesn't necessarily mean that it designates a file name below the _directory_ in the current directory setting, just that it's relative to some part of the current directory setting. Effectively, it's below the _root_ directory on the default drive.) Recall how URLs work: "xyz" and "/xyz" are both relative URL references. The first is relative to the "directory" portion of the base URL's pathname ("http://host/dir/x.html"; + "xyz" -> "http://host/dir/xyz";). The second is not relative to that "directory" portion, but is still relative to the scheme and authority portions of the base URL ("http://host/dir/x.html"; + "/xyz" -> "http://host/xyz";.) If we haven't diverged yet, then: - Do Windows (or Java?) API functions resolve pathnames differently than the behavior that I've seen in the DOS shell? E.g., if there's a function to resolve a pathname against a reference pathname (like Java's java.io.File.getCanonicalPath()), does it resolve things differently that I described above? - Does Ant treat Windows pathnames differently that Windows does? E.g., does Ant define "\xyz" as not being relative even though its meaning is relative to the current-drive portion of the current directory setting? Or does Ant not resolve Windows pathnames the same way Windows does? E.g., in a fileset, given a directory of "D:/", would a file pattern specification of "/x" not match "D:/x"? Thanks for bearing with this. Daniel -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
