@worley, nevertheless in some cases the results can be different: for example, if 'b' is directory that contains another symlinks "x" and "y" that point to some files, the result with -follow could be:

./b -- a
./b/x -- x1
./b/y -- y1

It should be so because there always exists hidden action  "process the directory content", and when a symlink is dereferenced to a directory, this action should be applied.

Yes, in case of -L option the behavior conforms the documentation. I just said that it seems not so useful (and maybe unexpected without careful investigation of the documentation).


But what about -follow option that is after -print and after -type test? For me, in such case the behavior of -print and -type contradicts the documentation.

"The -daystart, -follow and -regextype options  are  different  in  this  respect,  and  have an effect only on tests which appear later in the command line"

But currently -follow option has effect on the -type and -print which appear before it in the command line!


Certainly one can say that "-print" is not a test, but an action.  Actually, it doesn't matter. General respect for the position of -follow option (for actions as well as for tests) is desirable behavior that doesn't contradict the documentation and conforms the spirit and idea of the positional options.

Perhaps in case of -follow option (and without -L) the consistent way should be:
<pre>find . -type l -printf '%p -- %l\n' -follow</pre>
I.e. -type in place of -xtype  (for the moment it doesn't work).

Because the only options after -follow should consider the object of symlink, but all the options before -follow should consider the symlink itself.

Citation from the manual:

1. "any symbolic links appearing after -follow on the command line will be dereferenced, and those before it will not"

2. "Similarly, the -type predicate will always match against the type of the file that a symbolic link points to rather than the link itself."

So, -type before -follow should match against type of the file itself.