[build] allow ruby and python to run remote tests on windows

[titusfortner]

Situation

I noticed that Ruby can't run remote tests on Windows right now (server crashes) and I tracked it down to a real JDK symlink bug

• Java avoids this by running Grid tests in the same process
• Ruby and Python are currently skipping running Grid tests on Windows
• .NET runs Grid tests on Windows using System Java not Bazel Java (to be addressed later)
• JS doesn't run any tests on Windows

💥 What does this PR do?

Creates Ruby & Python workarounds to get the JDK home, resolve it via runfiles then get the real path of Java on Windows to avoid the symlink.

🔧 Implementation Notes

I didn't want to add new Windows Remote tests with this PR, but to show that the new code works: https://github.com/SeleniumHQ/selenium/actions/runs/26725902074/job/78761168501

Considered:

• Turning off symlinks - dramatically balloons the disk space to put everything in every directory
• Pointing at a non-Bazel JAVA_HOME on Windows (what .NET bindings do now) breaks hermeticity.
• Fixing the genrule at build time to emit a non-symlinked path — not possible; the symlink only exists in the per-test runfiles tree at runtime.

🤖 AI assistance

• No substantial AI assistance used
• AI assisted (complete below)
  • Tool(s): Claude Code
  • What was generated: `
  • I reviewed all AI output and can explain the change

💡 Additional Considerations

• .NET should add a new genrule to use hermetic java for tests

🔄 Types of changes

• Bug fix (backwards compatible)

[qodo-code-review]

Review Summary by Qodo

(Agentic_describe updated until commit 4fb048c)

Resolve JDK symlink paths for Windows remote tests

🐞 Bug fix

Walkthroughs

Description

• Resolve JDK symlink path on Windows for Ruby and Python
• Use os.path.realpath() in Python to resolve symlink
• Use File.realpath() in Ruby to resolve symlink
• Add platform check to apply fix only on Windows

Diagram

flowchart LR
  A["JDK Path from Runfiles"] --> B["Resolve via rlocation"]
  B --> C{"Windows Platform?"}
  C -->|Yes| D["Apply realpath Resolution"]
  C -->|No| E["Use Path As-Is"]
  D --> F["Java Path Ready"]
  E --> F

Loading

File Changes

1. py/conftest.py 🐞 Bug fix +7/-3

Add Windows JDK symlink resolution for Python

• Check for SE_BAZEL_JAVA_LOCATION environment variable before processing
• Apply os.path.realpath() to Java path on Windows platform
• Verify path exists before applying realpath resolution

py/conftest.py

---

2. rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb 🐞 Bug fix +2/-1

Add Windows JDK symlink resolution for Ruby

• Store resolved path in intermediate variable
• Apply File.realpath() on Windows platform only
• Conditionally resolve symlink based on platform and file existence

rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb

---

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (2) 📘 Rule violations (2)

1. Missing env var crashes 🐞 Bug ≡ Correctness

Description

In py/conftest.py, the remote server fixture concatenates a string with
os.environ.get("SE_BAZEL_JAVA_HOME") without checking for None, so an unset env var raises TypeError
before the try/except and aborts the session.

Code

py/conftest.py[569]

Evidence

The fixture concatenates a string with a possibly-missing env var outside the try/except, which
deterministically raises TypeError when the env var is absent.

py/conftest.py[566-572]

Agent prompt

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

### Issue description
`server()` builds a runfiles path via `"_main/" + os.environ.get("SE_BAZEL_JAVA_HOME")`. If the env var is unset, `os.environ.get(...)` returns `None` and Python raises `TypeError` during string concatenation. This happens **before** the `try:` block, so the exception is not caught and the whole test session fails.

### Issue Context
This code runs only for `--remote` (Grid) sessions. It is expected to work under Bazel, but it should also fail gracefully (or at least fail with a clear error) if the env var is missing.

### Fix Focus Areas
- py/conftest.py[566-582]

### Suggested change
- Read the env var into a local variable and guard it:
 - If unset: either (a) skip setting `server.java_path` (leaving default behavior) or (b) raise a clear error like `RuntimeError("SE_BAZEL_JAVA_HOME not set; cannot locate Bazel JDK")`.
- Ensure the `Rlocation(...)` call is only executed when you have a non-empty env var value.

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

---

2. SE_BAZEL_JAVA_LOCATION renamed 📘 Rule violation ⚙ Maintainability

Description

The PR replaces SE_BAZEL_JAVA_LOCATION with SE_BAZEL_JAVA_HOME without any backward-compatible
fallback, which can break existing test runners or scripts that still set the old variable. This
risks breaking consumers of this user-facing configuration surface.

Code

py/conftest.py[R569-572]

Evidence

PR Compliance ID 1 requires avoiding breaking changes to public/user-facing interfaces. The updated
Bazel test target sets only SE_BAZEL_JAVA_HOME, and the Python test harness reads only
SE_BAZEL_JAVA_HOME with no compatibility fallback for the previous SE_BAZEL_JAVA_LOCATION name.

AGENTS.md: Maintain API/ABI Compatibility for Public Interfaces
py/BUILD.bazel[814-821]
py/conftest.py[566-580]

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/conftest.py` now reads only `SE_BAZEL_JAVA_HOME`, but older invocations may still provide `SE_BAZEL_JAVA_LOCATION`. This creates a breaking change in configuration/API surface.

## Issue Context
The Bazel test rule now sets `SE_BAZEL_JAVA_HOME`, but external runners or pinned tooling may still set the old variable name.

## Fix Focus Areas
- py/conftest.py[566-580]
- py/BUILD.bazel[814-821]

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

---

3. WD_BAZEL_JAVA_LOCATION renamed 📘 Rule violation ⚙ Maintainability

Description

The PR replaces WD_BAZEL_JAVA_LOCATION with WD_BAZEL_JAVA_HOME without a backward-compatible
fallback, which can break existing scripts or Bazel invocations relying on the old variable name.
This is a potential breaking change for consumers of the Ruby test environment interface.

Code

rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[R137-140]

Evidence

PR Compliance ID 1 requires maintaining compatibility for public/user-facing interfaces. The Bazel
rule now exports only WD_BAZEL_JAVA_HOME, and the Ruby helper checks only for WD_BAZEL_JAVA_HOME
(no fallback to the old variable), which can break existing consumers still using
WD_BAZEL_JAVA_LOCATION.

AGENTS.md: Maintain API/ABI Compatibility for Public Interfaces
rb/spec/tests.bzl[222-232]
rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[136-149]

Agent prompt

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

## Issue description
The Ruby test environment now requires `WD_BAZEL_JAVA_HOME` and no longer accepts `WD_BAZEL_JAVA_LOCATION`, which can break existing invocations.

## Issue Context
`rb/spec/tests.bzl` sets the new variable name, but external consumers may still set the old one.

## Fix Focus Areas
- rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[136-149]
- rb/spec/tests.bzl[222-232]

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

---

4. Silent system Java fallback 🐞 Bug ☼ Reliability

Description

If resolving the Bazel JDK home fails in py/conftest.py, the code swallows the exception and leaves
Server.java_path unset; Server.start() then uses shutil.which("java"), which can bypass the intended
hermetic JDK and Windows realpath workaround.

Code

py/conftest.py[R570-582]

Evidence

py/conftest.py suppresses all exceptions during java path setup, and the Server implementation
explicitly falls back to PATH Java when java_path is unset.

py/conftest.py[554-582]
py/selenium/webdriver/remote/server.py[172-183]

Agent prompt

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

### Issue description
The Java-home resolution block catches `Exception` and does nothing. If `java_home_txt` can't be located/read, `Rlocation` returns None, or the constructed `bin/java(.exe)` path doesn't exist, the assignment to `server.java_path` is skipped. Downstream, `Server.start()` falls back to `shutil.which("java")`, which can select a system Java instead of the Bazel JDK.

### Issue Context
This PR’s goal is to reliably use Bazel’s JDK (and on Windows, its real path) for Grid tests. Silent fallback makes failures non-obvious and can reintroduce the original Windows crash if system Java points into the runfiles symlink farm.

### Fix Focus Areas
- py/conftest.py[570-582]

### Suggested change
- Narrow the exception handling and/or add explicit validation:
 - If `SE_BAZEL_JAVA_HOME` is set but `java_home_txt`/`java_home`/`server.java_path` cannot be resolved, raise a clear error (preferred for CI correctness), or at minimum log a warning with details.
 - Check `java_home_txt` and `java_home` for falsy values before proceeding.
 - Optionally assert that the final `server.java_path` exists before setting it (so failures are explicit).

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

---

Previous review results

Review updated until commit 4fb048c

Results up to commit 88061f6

🐞 Bugs (2) 📘 Rule violations (2) 📎 Requirement gaps (0)

1. Missing env var crashes 🐞 Bug ≡ Correctness

Description

In py/conftest.py, the remote server fixture concatenates a string with
os.environ.get("SE_BAZEL_JAVA_HOME") without checking for None, so an unset env var raises TypeError
before the try/except and aborts the session.

Code

py/conftest.py[569]

Evidence

The fixture concatenates a string with a possibly-missing env var outside the try/except, which
deterministically raises TypeError when the env var is absent.

py/conftest.py[566-572]

Agent prompt

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

### Issue description
`server()` builds a runfiles path via `"_main/" + os.environ.get("SE_BAZEL_JAVA_HOME")`. If the env var is unset, `os.environ.get(...)` returns `None` and Python raises `TypeError` during string concatenation. This happens **before** the `try:` block, so the exception is not caught and the whole test session fails.

### Issue Context
This code runs only for `--remote` (Grid) sessions. It is expected to work under Bazel, but it should also fail gracefully (or at least fail with a clear error) if the env var is missing.

### Fix Focus Areas
- py/conftest.py[566-582]

### Suggested change
- Read the env var into a local variable and guard it:
 - If unset: either (a) skip setting `server.java_path` (leaving default behavior) or (b) raise a clear error like `RuntimeError("SE_BAZEL_JAVA_HOME not set; cannot locate Bazel JDK")`.
- Ensure the `Rlocation(...)` call is only executed when you have a non-empty env var value.

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

---

2. SE_BAZEL_JAVA_LOCATION renamed 📘 Rule violation ⚙ Maintainability

Description

The PR replaces SE_BAZEL_JAVA_LOCATION with SE_BAZEL_JAVA_HOME without any backward-compatible
fallback, which can break existing test runners or scripts that still set the old variable. This
risks breaking consumers of this user-facing configuration surface.

Code

py/conftest.py[R569-572]

Evidence

PR Compliance ID 1 requires avoiding breaking changes to public/user-facing interfaces. The updated
Bazel test target sets only SE_BAZEL_JAVA_HOME, and the Python test harness reads only
SE_BAZEL_JAVA_HOME with no compatibility fallback for the previous SE_BAZEL_JAVA_LOCATION name.

AGENTS.md: Maintain API/ABI Compatibility for Public Interfaces
py/BUILD.bazel[814-821]
py/conftest.py[566-580]

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/conftest.py` now reads only `SE_BAZEL_JAVA_HOME`, but older invocations may still provide `SE_BAZEL_JAVA_LOCATION`. This creates a breaking change in configuration/API surface.

## Issue Context
The Bazel test rule now sets `SE_BAZEL_JAVA_HOME`, but external runners or pinned tooling may still set the old variable name.

## Fix Focus Areas
- py/conftest.py[566-580]
- py/BUILD.bazel[814-821]

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

---

3. WD_BAZEL_JAVA_LOCATION renamed 📘 Rule violation ⚙ Maintainability

Description

The PR replaces WD_BAZEL_JAVA_LOCATION with WD_BAZEL_JAVA_HOME without a backward-compatible
fallback, which can break existing scripts or Bazel invocations relying on the old variable name.
This is a potential breaking change for consumers of the Ruby test environment interface.

Code

rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[R137-140]

Evidence

PR Compliance ID 1 requires maintaining compatibility for public/user-facing interfaces. The Bazel
rule now exports only WD_BAZEL_JAVA_HOME, and the Ruby helper checks only for WD_BAZEL_JAVA_HOME
(no fallback to the old variable), which can break existing consumers still using
WD_BAZEL_JAVA_LOCATION.

AGENTS.md: Maintain API/ABI Compatibility for Public Interfaces
rb/spec/tests.bzl[222-232]
rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[136-149]

Agent prompt

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

## Issue description
The Ruby test environment now requires `WD_BAZEL_JAVA_HOME` and no longer accepts `WD_BAZEL_JAVA_LOCATION`, which can break existing invocations.

## Issue Context
`rb/spec/tests.bzl` sets the new variable name, but external consumers may still set the old one.

## Fix Focus Areas
- rb/spec/integration/selenium/webdriver/spec_support/test_environment.rb[136-149]
- rb/spec/tests.bzl[222-232]

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

---

4. Silent system Java fallback 🐞 Bug ☼ Reliability

Description

If resolving the Bazel JDK home fails in py/conftest.py, the code swallows the exception and leaves
Server.java_path unset; Server.start() then uses shutil.which("java"), which can bypass the intended
hermetic JDK and Windows realpath workaround.

Code

py/conftest.py[R570-582]

Evidence

py/conftest.py suppresses all exceptions during java path setup, and the Server implementation
explicitly falls back to PATH Java when java_path is unset.

py/conftest.py[554-582]
py/selenium/webdriver/remote/server.py[172-183]

Agent prompt

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

### Issue description
The Java-home resolution block catches `Exception` and does nothing. If `java_home_txt` can't be located/read, `Rlocation` returns None, or the constructed `bin/java(.exe)` path doesn't exist, the assignment to `server.java_path` is skipped. Downstream, `Server.start()` falls back to `shutil.which("java")`, which can select a system Java instead of the Bazel JDK.

### Issue Context
This PR’s goal is to reliably use Bazel’s JDK (and on Windows, its real path) for Grid tests. Silent fallback makes failures non-obvious and can reintroduce the original Windows crash if system Java points into the runfiles symlink farm.

### Fix Focus Areas
- py/conftest.py[570-582]

### Suggested change
- Narrow the exception handling and/or add explicit validation:
 - If `SE_BAZEL_JAVA_HOME` is set but `java_home_txt`/`java_home`/`server.java_path` cannot be resolved, raise a clear error (preferred for CI correctness), or at minimum log a warning with details.
 - Check `java_home_txt` and `java_home` for falsy values before proceeding.
 - Optionally assert that the final `server.java_path` exists before setting it (so failures are explicit).

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

---

[qodo-code-review]

Code review by qodo was updated up to the latest commit 4fb048c