Re: confusing error message from ln

[Paul Eggert]

address@hidden (Bob Proulx) writes:

> But the hard link case is much more complicated than before.  And
> unfortunately does not cover the main case.

I'm afraid it's the best we can do for ENOENT without issuing more
system calls.  If link(a,b) fails with ENOENT, it could be a problem
with either a or b.

> Worried that list is still not all encompassing and not all of them
> would be supported everywhere and that type of thing.

Except for EDQUOT, all the error numbers were in Unix version 7 and
have been required by POSIX for quite some time, so they should be
quite safe.  I will add this to system.h:

#ifndef EDQUOT
# define EDQUOT (-1)
#endif

to catch the oddball systems that lack disk-quota hooks.

I understand your point about weird systems, but it's just a
diagnostic that we're trying to improve the quality of, so there's no
possible conformance issue here -- it's just a UI issue.  So I'm still
inclined to try to do a better job with the hard-link diagnostic, even
if this sometimes means we might output a subset of the necessary info
on weird systems that we don't know about right now.