Fix Linux move_to_trash wrongly reporting files as not found

[akien-mga]

We can't rely on the error code from gio or kioclient5, in my rudimentary testing they return 1 for both missing files and other situations like not having a Trash can on the mounted volume.

I added path existence checks to all platforms as that seemed useful, even if it was wrongly implemented for Linux.
Edit: This would need more work, I left it out of this PR for now.

Fixes #79108.
Partially reverts #67158, but still fixes #67157.

CC @amoriqbal @romlok @starccy I only did basic tests on Linux, would appreciate you checking that it solves #79108.
You can test with CI artifacts once the build finishes.

[starccy]

Thank you for your work, I'll test it later.

[starccy]

I can confirm the error has disappeared on my system (Linux + Gnome).

[romlok]

The only Linux editor artifact I could see was the -mono version, which didn't work on my machine, but I was able to compile the branch locally and can confirm that this change fixes the issue for me too 👍

[YuriSizov]

Approving based on the comments above.

[starccy]

Oops, I missed a special use case.
When I create a broken symlink file, I cannot delete the file itself. And the "No such file or directory" error be presented.

if I get it correctly, the DirAccess::file_exists only checks the file it points to, but not for the symlink itself.

godot/drivers/unix/dir_access_unix.cpp

Line 75 in ef155c1

bool success = (stat(p_file.utf8().get_data(), &flags) == 0);

I was wondering why to use stat here instead of lstat, is this an expected behavior?

[akien-mga]

Thanks for testing further, that's indeed a regression compared to 4.0.3, where move_to_trash handles deleting symlinks fine on Linux.

Changing stat to lstat in DirAccessUnix could be considered but it's a very core change with a lot of implications, that would need to be assessed in depth.

For this PR I'll remove the check added in all platforms to handle missing files, and just let Linux behave like Windows and macOS: no validation.

---

On my system the PR in its current state (a6e75f3) prints these errors when trying to remove a file which doesn't exist:

gio: file:///home/akien/Test/test4: Error trashing file /home/akien/Test/test4: No such file or directory
kf.kio.workers.trash: Directory "/media/windows/.Trash-1000" exists but didn't pass the security checks, can't use it
kioclient5: Access denied to trash:/http:⁄⁄test4⁄.
ERROR: Could not create child process: gvfs-trash
   at: execute (./drivers/unix/os_unix.cpp:541)
ERROR: Can't rename file "test4" to "/test4"
   at: move_to_trash (platform/linuxbsd/os_linuxbsd.cpp:1075)

It's bad, but it's the same as in 4.0.3 so this still fixes the regression and should be suitable for a 4.1.1 cherry-pick.

The whole move_to_trash functionality could use a good cleanup on all platforms to properly handle validating input, and make the Linux one a bit sturdier as the current code seems a bit brittle to me.

[starccy]

Yes, the error messages from gio/kioclient5 are annoying, but thankfully the method still works as properly : )

Changing stat to lstat in DirAccessUnix could be considered but it's a very core change with a lot of implications, that would need to be assessed in depth.

In fact, some languages like Python, have separate APIs to let users decide to choose the one they think is appropriate: os.path.exist and os.path.lexists, which can be imitated in gdscript to avoid breaking changes as much as possible. AFAIK, the gdscript users have no chance to determine if a file is a symlink. They might never or hard to check existence for a broken symbolic link.

[YuriSizov]

Cherry-picked for 4.1.1.