Path-Tiny-0.028 Path::Tiny on win32 turns C:/ or C:\ into C: / absolute / realpath are broken

[Anonymous Monk]

Path-Tiny-0.028 Path::Tiny on win32 turns C:/ or C:\ into C: but those aren't equivalent (C: is chdir C: && echo %cd% )

This breaks absolute and realpath (resolve)

$ perl -MPath::Tiny -le "  print path( 'C:/')->realpath  "
C:/Documents and Settings/username

$ perl -MPath::Tiny -le "  print path( 'D:/')->realpath  "
D:/dev/misc

$ perl -MPath::Tiny -le "  print path( 'C:/')->absolute->realpath  "
D:/dev/misc/C:: No such file or directory at C:/perl/site/lib/Path/Tin
+y.pm line 449.

$ mversion Path::Tiny
0.028

Editing sub path to comment out the following line seems to fix it :)

$path =~ s{/$}{} if $path ne "/"; # hack to make splitpath give us a b
+asename

There are 2 extra failures as a result of this patch (meaning its working)

not ok 55 - path('A:/')
#   Failed test 'path('A:/')'
#   at t/zzz-spec.t line 200.
#          got: 'A:/'
#     expected: 'A:'

not ok 67 - path('A:/')
#   Failed test 'path('A:/')'
#   at t/zzz-spec.t line 200.
#          got: 'A:/'
#     expected: 'A:'

Does this break on linux?

[tobyink]

It doesn't break the test suite, but I don't know whether it breaks any usage patterns that aren't tested by the test suite.

Report your problem and suggested solution to the issue tracker - David Golden is pretty quick on the ball for Path::Tiny issues.

[vsespb]

Does this break on linux?

No

[xdg]

FYI: This has been opened as issue #49.

Does this break on Linux

It should. So that's a test suite coverage failure.

I think I have a fix, but I have a broader question:

Is there any reason to allow a raw volume like "C:" (no trailing slash) or "\\server\share" as a Path::Tiny "path"? Or should I be fixing those to have a trailing slash so they are properly understood as referencing the root directory of those volumes?

[Corion]

I would say that there is a misconception. C: does not reference the root directory of a volume. It references the current working directory of that volume.

You should either prohibit this or handle it properly, not silently convert to something else, especially not a root directory. I think the failure mode of converting C: to C:/ is a very bad one:

my $p= path("C:");
unlink $_ for $p->children;

For UNC paths, I see no such bad failure mode of appending a trailing directory separator.

[xdg]

C: does not reference the root directory of a volume

I understand that. What I'm trying to get guidance on is how likely it is that people would use "C:" and not mean the root directory.

If I understand you correctly, what probably needs to happen is to immediately turn C: into the absolute path for cwd on that volume via getdcwd.

Is that right?