Need more tests for `W`/`Z` suffixed `std.fs` functions

[squeek502]

In light of #16736 being closed, this:

Note also that pretty much all Absolute functions have poor test coverage to the point that some will give a compile error if anyone tried to actually use them (e.g. on Windows anything that tries to call the no-longer-existing os.windows.cStrToWin32PrefixedFileW function [these functions have never worked since their introduction in #5879 which ended up removing cStrToWin32PrefixedFileW before it was merged but left some calls of it around])

needs to be addressed by actually fixing and improving the test coverage of the Z/W suffixed functions.

Relevant test file:

https://github.com/ziglang/zig/blob/master/lib/std/fs/test.zig

[matu3ba]

@squeek502 Can you be more specific how those tests are insufficient? Except deleteTreeAbsolute each fn is used at least once in https://github.com/ziglang/zig/blob/master/lib/std/fs/test.zig as of bf5ab54.

• Do you mean that it does not test all file path encodings (ntdll networkpaths and regular ones on Windows) ?
• Can you tldr; what the suggested strategy was regarding the absolute functions to keep things portable? Unfortunately I don't know a wrapup and the comments do not mention in what ways absolute paths can or can not be used or relied upon.
• What should code do, which relies on some absolute paths of the platform for security guarantees to prevent attack vectors ?

Summary of current state:

general

• accessAbsolute: test "accessAbsolute" (used in test ". and .. in absolute functions")
• renameAbsolute: test "renameAbsolute" (used in test ". and .. in absolute functions")

identically named

• openDirAbsolute: test "openDirAbsolute"
• openFileAbsolute (used in test ". and .. in absolute functions")
• deleteFileAbsolute (used in test "file operations on directories", test "open file with exclusive nonblocking lock twice (absolute paths)")
• deleteDirAbsolute (used in test "directory operations on files", test ". and .. in absolute functions")
• deleteTreeAbsolute: not tested

weirdly differently named

• createFileAbsolute (used in test "file operations on directories", test "open file with exclusive nonblocking lock twice (absolute paths)")
• makeDirAbsolute (used in test "directory operations on files", test ". and .. in absolute functions")

differently named

• copyFileAbsolute (used in test ". and .. in absolute functions")
• updateFileAbsolute (used in test ". and .. in absolute functions")
• readLinkAbsolute: test "readLinkAbsolute"
• symlinkAbsolute (used in test "readLinkAbsolute")

[squeek502]

Sorry, I had things a bit mixed up.

The real issue is the W/Z suffixed versions of the Absolute functions. For example, readLinkAbsoluteZ and symLinkAbsoluteZ are compile errors on Windows currently. This lack of testing of the W/Z suffixed versions probably also applies to the Dir functions as well.