Fix File.cp_r/3 infinite loop when copying into subdirectory of source

[pnezis]

When destination is a subdirectory of source, cp_r would enter an infinite loop because the newly created destination directory would be included in the file listing during recursion. This can cause disk exhaustion if large files are present under source.

This adds a validation check that returns {:error, :einval, destination} when the expanded destination path starts with the expanded source path followed by a path separator.

Also document typical error reasons including the new :einval case.

Simple reproduction:

iex> File.mkdir_p!("/tmp/src/subdir")
iex> File.write!("/tmp/src/file.txt", "hello")
iex> File.cp_r("/tmp/src", "/tmp/src/subdir/dest")
{:error, :enametoolong,
"/tmp/src/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/...file.txt"}

Symlinks

Notice symlinks are not resolved by Path.expand, so the issue is still there if you are using a symlinked subdir:

iex> File.mkdir_p!("/tmp/src/subdir")
iex> File.write!("/tmp/src/file.txt", "hello")
iex> File.mkdir_p!("/tmp/outside")
iex> :file.make_symlink("/tmp/src/subdir", "/tmp/outside/link")
iex> File.cp_r("/tmp/src", "/tmp/outside/link/dest")
{:error, :enametoolong,
"/tmp/outside/link/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/subdir/dest/...dest"}

• I could not find a stdlib elixir/erlang function that resolves symlinks, should we add a realpath ?
• I have found another issue with infinite symlink cycles, that may handle this. Will handle in another PR.

[pnezis]

Looks that :einval is the most appropriate here, any better suggestion?

[pnezis]

A Path.parent? would be useful here.

[josevalim]

Giveni the symlinks behaviour, I am not sure if it is something we should try to solve codewise or if we should just warn users of the problematic behaviour. In any case, if we do want to check for parent/child relationship, the best is to split call Path.split and then List.start_with?

[pnezis]

Pushed the fix with Path.split and List.starts_with, I think it is better to be defensive than just warn, since you can easily end up with a filled disk if by mistake you copy a big dir, to a subdir of it.

[josevalim]

Tests are failing :)

[pnezis]

Fixed, there was a test for copying src to src, that was failing with the List.starts_with

[josevalim]

💚 💙 💜 💛 ❤️