On 21/10/2019 11:11, Andreas Schwab wrote:
On Okt 21 2019, Marco Ippolito wrote:
In the GNU Bash 5.0 man page from 7 December 2018 the -e test is
documented as such:
-e file
True if file exists.
When "file" is a symlink name to a non-existing target, the -e test fails,
and this may be surprising from just reading the documentation.
It also says:
Unless otherwise specified, primaries that operate on files follow sym-
bolic links and operate on the target of the link, rather than the link
itself.
Andreas.
I still personally find this overloading of the term "file" to refer to:
(1) a generic operand of those primaries we'll call "file", be it an
actual file or something else
(2) an actual file
somewhat unnecessarily ambiguous compared to the POSIX version.
As a minor point, when I read the POSIX's wording of "resolves to an
existing directory entry", I find it intuitive, possibly from
familiarity with other Computer Science contexts like DNS in which
look-ups happen in a chain, that the word "resolves" might entail
recursion, as in the case I highlighted in code, where a link's target
was another symlink. When I read "follow", especially paired with the
singular "the target of the link" (not "of the links" or "of the link
chain" or similia - and by the way, why use "link" here and not
"symlink"? to give the name "link" to a "chain of symlinks"?), I am left
wondering for a moment if that implies recursion or not, which is
something I did not, subjectively, wonder about when reading the
alternative word "resolves". After all, the man page for readlink uses
the word "follow" to read a link's target but it makes the point of
being explicit about recursive application for the canonicalisation
options, as if it weren't automatic to think of "following a link" as
"following it... until the end":
-f, --canonicalize
canonicalize by following every symlink in every
component of the given name recursively; all but the last component must
exist
-e, --canonicalize-existing
canonicalize by following every symlink in every
component of the given name recursively, all components must exist
-m, --canonicalize-missing
canonicalize by following every symlink in every
component of the given name recursively, without requirements on
components existence
If following is meant to be understood as "always to the end of the
symlink chain", it seems redundant to say to follow _every_ symlink and
to do so _recursively_.
So, in summary:
- file still looks less precise than pathname to me
- (minor point) resolution seems like an elegant term to copy from POSIX
to indicate the recursive process, whereby to "follow", especially
paired with a singular noun (i.e. link), might be more confusing to some
