Issue #7106 has been updated by luislavena (Luis Lavena).

Status changed from Assigned to Feedback

=begin

I've combined your patch and Usa's Windows modification (plus other misses) and put it here:

https://gist.github.com/4022459

Also, I've created the following files on a drive:

* foo: directory
* READ: a real file
  
And created the following type of links:

* bar: symlinkd (directory symlink) of foo
* a-link: symlink of READ
* HARD: hardlink of READ
* junc: a junction point of foo
  
 V:\>dir
 Volume in drive V is RAMDISK
 Volume Serial Number is 9A3E-37F6
 
 Directory of V:\
 
 06/11/2012  12:57 a.m.    <SYMLINK>      a-link [READ]
 06/11/2012  12:54 a.m.    <SYMLINKD>     bar [foo]
 05/11/2012  09:25 p.m.    <DIR>          foo
 06/11/2012  12:56 a.m.                 7 HARD
 06/11/2012  01:02 a.m.    <JUNCTION>     junc [V:\foo]
 06/11/2012  12:56 a.m.                 7 READ
 06/11/2012  12:49 a.m.    <DIR>          ruby2
 
As you can see from the output, all have different timestamps

Now, from IRB:

 V:\>irb
 irb(main):001:0> File.mtime "foo"
 => 2012-11-05 21:25:54 -0300
 irb(main):002:0> File.mtime "bar"
 => 2012-11-06 00:54:22 -0300
 irb(main):003:0> File.mtime "junc"
 => 2012-11-06 01:02:03 -0300
 
On Windows, access of modification time gets the link modification time and not the target.

Same goes for atime and ctime:

 irb(main):004:0> ["foo", "bar", "junc"].each { |folder| puts File.atime(folder) }
 2012-11-05 21:25:54 -0300
 2012-11-06 00:54:22 -0300
 2012-11-06 01:02:03 -0300
 
 irb(main):006:0> ["foo", "bar", "junc"].each { |folder| puts File.ctime(folder) }; nil
 2012-11-05 21:23:45 -0300
 2012-11-06 00:54:22 -0300
 2012-11-06 01:02:03 -0300
 
Now, doing (({File.utime})) on the symlinkd modifies the target:

 irb(main):016:0> a = File.mtime("foo")
 => 2012-11-07 00:54:22 -0300
 irb(main):017:0> a = File.atime("bar")
 => 2012-11-06 00:54:22 -0300
 irb(main):018:0> File.utime(a, a, "bar")
 => 1
 irb(main):019:0> a = File.atime("bar")
 => 2012-11-06 00:54:22 -0300
 irb(main):020:0> a = File.atime("foo")
 => 2012-11-06 00:54:22 -0300
 
But not the symlink, which is expected.

Problem is now that the code around (({rb_file_s_lutime})) avoids it getting defined on Windows, even when functions around it were faked.
(it checks for (({HAVE_LUTIMES})) and that doesn't exists on Windows)

I did a minor tweak and got that passing (is included in my patch), and now:

 V:\>irb
 irb(main):001:0> a = File.atime("bar")
 => 2012-11-06 00:54:22 -0300
 irb(main):002:0> b = a + 3600
 => 2012-11-06 01:54:22 -0300
 irb(main):003:0> File.lutime(b, b, "bar")
 => 1
 irb(main):004:0> File.atime("bar")
 => 2012-11-06 01:54:22 -0300
 irb(main):005:0> File.atime("foo")
 => 2012-11-06 00:54:22 -0300
 
Symlinks can only be created by administrators, but modifying the timestamps (access, modification) is totally possible.

It also works transparently with junction points (which normal users can create too):

 irb(main):008:0> File.atime("junc")
 => 2012-11-06 01:02:03 -0300
 irb(main):009:0> File.atime("foo")
 => 2012-11-06 00:54:22 -0300
 irb(main):010:0> File.lutime(b, b, "junc")
 => 1
 irb(main):011:0> File.atime("foo")
 => 2012-11-06 00:54:22 -0300
 irb(main):012:0> File.atime("junc")
 => 2012-11-06 01:54:22 -0300
 
And file symlinks (previous were directory symlinks, are two different kind of links):

 irb(main):018:0> File.atime("READ")
 => 2012-11-06 00:56:43 -0300
 irb(main):019:0> File.atime("a-link")
 => 2012-11-06 00:57:07 -0300
 irb(main):020:0> c = _ + (4 * 3600)
 => 2012-11-06 04:57:07 -0300
 irb(main):021:0> File.lutime(c, c, "a-link")
 => 1
 irb(main):022:0> File.atime("READ")
 => 2012-11-06 00:56:43 -0300
 irb(main):023:0> File.atime("a-link")
 => 2012-11-06 04:57:07 -0300
 
The only one that has no effect are hardlinks, but because the file itself is the same (which is expected)

 irb(main):024:0> File.atime("READ")
 => 2012-11-06 00:56:43 -0300
 irb(main):025:0> File.atime("HARD")
 => 2012-11-06 00:56:43 -0300
 irb(main):026:0> d = _ + (9 * 3600)
 => 2012-11-06 09:56:43 -0300
 irb(main):027:0> File.lutime(d, d, "HARD")
 => 1
 irb(main):028:0> File.atime("READ")
 => 2012-11-06 09:56:43 -0300
 irb(main):029:0> File.atime("HARD")
 => 2012-11-06 09:56:43 -0300
 
All this was performed with a normal command prompt (without admin rights)

Now, (({File.lutime})) seems to work properly, but the modification to (({FileUtils})) might not work.

As you see, FileUtils relies on (({File.symlink?})) to combine with (({:nofollow})) option and then use either (({lutime})) or (({utime})).

Problem is, (({File.symlink?})) always return false on Windows:

 irb(main):001:0> File.symlink?("foo")
 => false
 irb(main):002:0> File.symlink?("bar")
 => false
 irb(main):003:0> File.symlink?("a-link")
 => false
 irb(main):004:0> File.symlink?("HARD")
 => false
 irb(main):005:0> File.symlink?("junc")
 
While current Ruby can create hardlinks, it cannot determine a symlink (the code returns directly (({Qfalse})).

I think that is the final part to get your code into Ruby (at least working on Windows).

Thank you.
=end
----------------------------------------
Feature #7106: FileUtils.touch should allow touching the symlink itself rather than the file the link points to
https://bugs.ruby-lang.org/issues/7106#change-32517

Author: cirrusthinking (Alessandro Diaferia)
Status: Feedback
Priority: Normal
Assignee: luislavena (Luis Lavena)
Category: core
Target version: 2.0.0


Most systems support POSIX lutimes sys call that allows a symlink to be touched. I'd like to see this feature implemented.
Unfortunately I'd like to see this feature on Solaris as soon as possible but I cannot see lutimes syscall supported there.


-- 
http://bugs.ruby-lang.org/