Pike v7.4 release 31 running Hilfe v3.5 (Incremental Pike Frontend)

file_stat("C:");

(1) Result: 0

file_stat("C:/");

(2) Result: 0

file_stat("C:\");

(3) Result: 0

/ Martin Nilsson (saturator)

Previous text:

2004-02-12 10:55: Subject: NT problem.

---

Does the same apply to "C:\"? I think the handling of "X:" without a following \ is poor in multiple places.

/ Marcus Comstedt (ACROSS) (Hail Ilpalazzo!)

I believe it should. Otherwise you would have to add these kinds of kludges whenever you stumble upon problems:

Stdio.Stat st = file_stat(path)

#ifdef __NT__ if(!st && get_dir(path)) st = Stdio.Stat( ({ 040777, -2, 0,0,0,0,0 }) ); #endif

if(st->isdir) recurse(path); else do_stuff(path);

Is there a good reason why this shouldn't work without the kludge?

/ Martin Nilsson (saturator)

Previous text:

2004-02-12 17:50: Subject: NT problem.

---

Can NT produce something stat-like for a unit letter? If it can't, I'm not sure that Pike should fake it.

/ Martin Stjernholm, Roxen IS

I think that everything that exists should return a non-zero result from file_stat. I have nothing against putting in methods in the stat object to differentiate between what we do know and what we don't. Let zero_type(st->mtime) be true?

/ Martin Nilsson (saturator)

Previous text:

2004-02-12 22:53: Subject: NT problem.

---

There are already other pecularities between NT and Unix file systems that require similar preprocessor stuff. The Stdio API should only try to hide such things as far as can be done without producing false pretenses. Returning nonzero if the drive letter exists is not disinformation, but putting arbitrary timestamps etc into a Stat struct is.

As for your example, you should use Stdio.is_dir instead. Stdio.exist and Stdio.is_dir should return nonzero in that case and I also consider it a bug that they don't.

/ Martin Stjernholm, Roxen IS