Bug 6068 – std.path has some issues for Windows user

getExt, getName and join doesn't support Windows's altsep, look at pull for D2's phobos: https://github.com/D-Programming-Language/phobos/pull/63 Confusing (for me) behaviour of functions: 1. isabs - from http://msdn.microsoft.com/en-us/library/aa365247%28v=vs.85%29.aspx : "A single backslash, for example, "\directory" or "\file.txt". This is also referred to as an absolute path." and in .NET Framework Path.IsPathRooted returns true for such case, e.g. But I agree with phobos's way, and this behaviour is clearly written in documentation. Have it your way. But it affects next function. 2. join - if second path is "drive relative rooted": assert( join(`a`, `\b`) == `\b` ); //good (like .NET Framework) assert( join(`c:a`, `\b`) == `c:a\b` ); //why not `c:\b`? assert( join(`c:\a`, `\b`) == `c:\a\b` ); //why not `c:\b`? .NET Framework's Path.Combine for all cases shows `\b` (because `\b` is rooted for it). Possible solutions: 1. We don't want to know about Microsoft's stupid "drive relative rooted" paths, let it be so: assert( join(`a`, `\b`) == `a\b` ); assert( join(`c:a`, `\b`) == `c:a\b` ); assert( join(`c:\a`, `\b`) == `c:\a\b` ); 2. We know about it: assert( join(`a`, `\b`) == `\b` ); assert( join(`c:a`, `\b`) == `c:\b` ); assert( join(`c:\a`, `\b`) == `c:\b` );

I would point out that Lars has revamped std.path, and his changes are likely to be up for review sometime soon (I believe that dsimcha's TempAlloc stuff is going to be reviewed first). And that's likely to change the behavior of some of std.path's stuff. Your issues here may be totally fixed in the code that Lars has done. So, I'm reluctant to spend time trying to fix what's currently there. As long as the pull request is good, it'll probably get merged in, but fixing the other issues should probably be left to Lars' changes, since I fully expect some version of that to get through the review process. We should keep these issues in mind however when Lars' changes are reviewed so that new code handles them in whatever the best way is.

Marking this as "won't fix", since the functions in question will soon be removed from Phobos.