Re: [Chicken-hackers] [PATCH] handle EOVERFLOW and ENOTDIR gracefully in file/directory-exists?

[Alan Post]

On Sat, Oct 08, 2011 at 04:11:47PM -0400, John Cowan wrote:
> Christian Kellermann scripsit:
> 
> > Wouldn't the right answer be in this case "We don't know whether this
> > file exists, because some Bad Thing(tm) happened while trying to find
> > out" instead of yes or no?
> 
> No.  If ENAMETOOLONG is returned, the file can't possibly exist (because
> the filesystem has no way to represent it), so it's always correct to
> return #f in that case.
> 
> > Depending on my use case I might be screwed either way if file-exists?
> > returns a valid answer.
> 
> Only in the sense that if it returns #f you may then try to create it,
> which will fail with another ENAMETOOLONG (which in this case should raise
> a Scheme exception).  So you have to be prepared for create to fail even
> if file-exists? returns #f (as in many other cases: permissions, e.g.)
> 

I'll add that in almost every case, checking for a file's existence
before trying to create it is the wrong thing to do, as it introuces
a race condition, (the file could exist and then be deleted before
you get around to doing the thing you were going to do with the
existent file.) and you should just go ahead and create the thing,
dealing with the error there.

-Alan
-- 
.i ma'a lo bradi cu penmi gi'e du