Fix py:local_dev rake task

[cgoldberg]

💥 What does this PR do?

This PR fixes the py:local_dev rake task so it deletes BiDi files from the source tree before copying them from bazel output. Without this, the task will fail since the files are read-only.

It also fixes the macos directory name so this works on Mac.

🔄 Types of changes

• Bug fix (backwards compatible)
• Development

[qodo-code-review]

Review Summary by Qodo

Fix py:local_dev rake task for BiDi file handling and macOS

🐞 Bug fix

Walkthroughs

Description

• Fix py:local_dev rake task to handle read-only BiDi files
• Delete destination files before copying to prevent permission errors
• Correct macOS directory name from 'mac' to 'macos'
• Simplify devtools cleanup by removing unnecessary pre-deletion

File Changes

1. rake_tasks/python.rake 🐞 Bug fix +4/-3

Fix BiDi file handling and macOS directory name

• Remove pre-deletion of devtools directory in copy_all branch
• Add explicit file deletion before copying BiDi files to handle read-only files
• Fix directory name from 'common/mac' to 'common/macos'
• Refactor to use dest variable for cleaner code

rake_tasks/python.rake

---

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (1)   📘 Rule violations (0)   📎 Requirement gaps (0)

🐞\ ≡ Correctness (1)

1. Stale BiDi files persist 🐞 ≡

Description

py:local_dev copies BiDi files individually without clearing py/selenium/webdriver/common/bidi,
so previously-generated BiDi modules that are no longer present in the Bazel output can remain
importable. This can cause local development/tests to run against stale BiDi code after generator
outputs change.

Code

rake_tasks/python.rake[R62-75]

    bidi_src = "#{bazel_bin}/common/bidi"

Evidence

The BiDi section only iterates entries currently present in bazel-bin/.../common/bidi and copies
them into py/.../common/bidi but never removes destination files that are absent from the source;
in contrast, other generated directories are fully replaced via rm_rf(dest) before copying. Since
BiDi outputs are explicitly declared/generated as a fixed set of files, any change to that set
(e.g., removing/renaming a module) will leave the old file behind in the source tree unless the
destination directory is cleared first.

rake_tasks/python.rake[62-75]
rake_tasks/python.rake[77-84]
py/private/generate_bidi.bzl[12-38]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`py:local_dev` updates `common/bidi` by copying only the files that currently exist in Bazel output, but it does not remove files that were generated previously and are no longer produced. This can leave stale modules in `py/selenium/webdriver/common/bidi`.

### Issue Context
Other generated directories (`common/devtools`, `common/linux`, etc.) are handled as full replacements (`rm_rf` then `cp_r`), which avoids stale files.

### Fix Focus Areas
- rake_tasks/python.rake[62-75]

### Proposed fix
- Before copying BiDi outputs, remove `bidi_dest` entirely (e.g., `FileUtils.rm_rf(bidi_dest)`) and then copy the whole directory from `bidi_src`.
- Keep the read-only overwrite fix by ensuring the destination is recreated (e.g., `mkdir_p`) after removal.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. copy_all leaves stale devtools ☑ 🐞 ≡

Description

In copy_all mode, py:local_dev no longer removes py/selenium/webdriver/common/devtools before
copying, so versioned DevTools protocol directories that are no longer generated can remain in the
source tree. This can cause local development to import outdated protocol bindings.

Code

rake_tasks/python.rake[R58-61]

  copy_all = arguments[:all] == 'all'
  if copy_all
-    FileUtils.rm_rf("#{lib_path}/common/devtools")
    FileUtils.cp_r("#{bazel_bin}/.", lib_path, remove_destination: true)
  else

Evidence

The copy_all branch now performs a bulk cp_r into py/selenium/webdriver without the previous
explicit cleanup of common/devtools, while the non-all path explicitly rm_rfs each generated
directory before copying (including common/devtools). DevTools outputs are generated into
versioned output directories (e.g., .../common/devtools/<version>), so the set of version
directories can change; without an explicit cleanup step, old version directories can remain in the
source tree across runs/branch switches.

rake_tasks/python.rake[58-61]
rake_tasks/python.rake[77-84]
py/BUILD.bazel[606-613]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`py:local_dev all` can leave stale `common/devtools` content in the source tree because it no longer clears that directory before copying, even though DevTools outputs are versioned and can change over time.

### Issue Context
The non-all path already fully replaces generated directories via `rm_rf(dest)` before copying; `copy_all` should avoid leaving old generated versions around for consistency.

### Fix Focus Areas
- rake_tasks/python.rake[58-61]

### Proposed fix
- Reintroduce an explicit `FileUtils.rm_rf("#{lib_path}/common/devtools")` before the `cp_r` in the `copy_all` branch (and consider similar cleanup for other versioned/generated subtrees if needed).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more

[cgoldberg]

The Formatter CI job failed because of:

Error in download: java.io.IOException: Error downloading [https://github.com/aspect-build/bsdtar-prebuilt/releases/download/v3.8.1/tar_linux_amd64] to /home/runner/.bazel/external/aspect_bazel_lib++toolchains+bsd_tar_linux_amd64/tar: GET returned 504 Gateway Time-out

I ran formatting locally and everything passed