This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Problem with stat

Eric Blake <> writes:

> Matthew Woehlke <mw_triad <at>> writes:
>> >> "If the named file is a symbolic link, the stat() function shall
>> >> continue pathname resolution using the contents of the symbolic
>> >> link, and shall return information pertaining to the resulting file
>> >> if the file exists."
>> > 
>> > This says nothing about what is returned if the file does not exist.
>>      Upon successful completion, 0 shall be returned. Otherwise, -1 shall
>>      be returned and errno set to indicate the error.
> Matthew is right.

I have no doubt that he is right about what programs do.  I am just
pointing out that it involved an advanced degree in lawyering and
sophistry to extract this kind of information from the standard.
Which makes the standard badly worded, at the very least.

> And if that isn't enough to convince you, you should also read POSIX
> XBD 4.11, which talks all about pathname resolution.

Yes, that is more clear.  But it does not help one standard that a
different standard is more careful in its wordings.

> Since stat() is not one of the special functions that operates on
> symlinks, it MUST resolve whatever the link points to, and if the
> link is broken, then it must behave the same way it does for any
> other resolution failure, ie. in this case, fail with ENOENT.  In
> short, it is IMPOSSIBLE for stat() to give you a struct stat that
> passes S_IFLNK on a compliant platform.  Give up your argument now.
> There are enough of us on this list that know POSIX and SUSv3 that
> you will not win.

Whatever.  I fail to see that the first wording of the standard is so
unambiguous that I would bet my life on S_ISLNK never returning true
on a dead link accessed with stat, when the implementor was of the
opinion of following the standard's wordings.

In short: I was not talking about what behavior to _implement_ but
rather what behavior to _expect_.  And given the fuzzy wording, I find
the intent to be reasonably well hidden here.

David Kastrup

Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]