On Wed, 24 May 2023 14:18:02 GMT, Alexey Ivanov <[email protected]> wrote:

>> `Files.isSymbolicLink`  takes path as argument to check if its a symbolic 
>> link. For certain windows specific files as explained the path won't be 
>> exact path, rather a hex values. So directly validating with path is causing 
>> an exception. So `fsv.isLink()` is used as a first level validation which 
>> uses File to check if its valid. I didn't find any other alternatives to 
>> validate the file path, so used `fsv.isLink` as first level validation along 
>> with Symbolic Link check.
>
> I understand that, virtual folders don't have a filesystem path, this is why 
> `fsv.isFileSystem(f)` should handle this situation. The implementation of 
> `FileSystemView.isFileSystem` explicitly excludes such files which are both 
> `isLink` and `isDirectory`:
> https://github.com/openjdk/jdk/blob/d7245f70e7bac1236bbcdcd9b25346ca22ab8bb2/src/java.desktop/share/classes/javax/swing/filechooser/FileSystemView.java#L382-L388
> 
> Let's look deeper:
> https://github.com/openjdk/jdk/blob/d7245f70e7bac1236bbcdcd9b25346ca22ab8bb2/src/java.desktop/windows/classes/sun/awt/shell/Win32ShellFolder2.java#L591-L597
> 
> https://github.com/openjdk/jdk/blob/d7245f70e7bac1236bbcdcd9b25346ca22ab8bb2/src/java.desktop/windows/classes/sun/awt/shell/Win32ShellFolder2.java#L860-L866
> 
> The `hasAttribute` method uses `getAttributes0` which calls 
> [`IShellFolder::GetAttributesOf`](https://learn.microsoft.com/en-us/windows/win32/api/shobjidl_core/nf-shobjidl_core-ishellfolder-getattributesof)
>  which returns 
> [`SFGAO`](https://learn.microsoft.com/en-us/windows/win32/shell/sfgao). The 
> attribute `SFGAO_LINK` which corresponds to `ATTRIB_LINK` in Java is 
> described as _“The specified items are shortcuts.”_ That is it is a shell 
> folder shortcut with `.lnk` file extension, which used to be the case.
> 
> However, now this attribute is set for both [_symbolic 
> links_](https://learn.microsoft.com/en-us/windows/win32/fileio/creating-symbolic-links)
>  and 
> [_junctions_](https://learn.microsoft.com/en-us/windows/win32/fileio/hard-links-and-junctions#junctions).
> 
> Here's what different methods return:
> 
> file: C:\test\symbolic-links\jdk - junction
>       isFileSystem   false
>       isLink         true
>       isDirectory    true
>       isSymbolicLink false
> Absolute Path : null
> 
> file: C:\test\symbolic-links\jdk - symbolic
>       isFileSystem   false
>       isLink         true
>       isDirectory    true
>       isSymbolicLink true
> Absolute Path : C:\test\symbolic-links\jdk - symbolic
> 
> file: C:\test\symbolic-links\jdk - shortcut.lnk
>       isFileSystem   false
>       isLink         true
>       isDirectory    true
>       isSymbolicLink false
> Absolute Path : null
> 
> 
> In the three cases above, `isFileSystem` returns `false` even though all the 
> three objects point to a filesystem directory, the first two are fully 
> transparent for apps whereas the `.lnk` shortcut needs special handling.
> 
> If `JFileChooser` is in directory-only mode, selecting a junction or `.lnk` 
> file results in selected file being to `null`; symbolic links are supported 
> with your fix.
> 
> So, **I propose mod...

Handling `.lnk` files on Windows doesn't seem right: `JFileChooser` allows 
selecting such a file.

* If it's a shortcut to a directory, clicking Open in the file chooser or 
double-clicking it in the list navigates to the target.
* If it's a shortcut to a regular file, `JFileChooser` returns this file which 
isn't helpful, it's not what the user usually wants, the target of the shortcut 
should be returned. This is what happens if the user selects a shortcut file 
(`.lnk`) in the native file chooser.

In addition to that, the shortcuts are filtered based on the file types. For 
example, if select the *Open* command in Notepad, the file chooser shows only 
shortcuts to `.txt` files (as well as `.txt` files), it hides shortcuts to 
other file types.

But `JFileChooser` doesn't show `.lnk` files at all if an extension filter is 
set.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/13998#discussion_r1204423861

Reply via email to