Add platform-specific variable overrides

[james-w]

Summary

This PR adds platform-specific variable overrides to handle cross-platform differences in commands and configurations.

Feature

Enable platform-specific variable overrides in both [globals] and target-level variables sections to support cross-platform commands.

Syntax:

[globals]
tool = "./tool"

[globals.platform.windows]
tool = "./tool.exe"

[globals.platform.linux]
tool = "./tool"

[globals.platform.macos]
tool = "./tool"

Key Features:

• Reserved word: platform cannot be used as a variable name
• Unknown platforms cause deserialization errors
• Backward compatible: existing configs work unchanged
• Early resolution: platform overrides resolved during config parsing
• Empty values supported: allows clearing variables on specific platforms
• Compile-time platform detection with error on unsupported platforms

Example usage:
See pls.toml for a practical example with the local-install command that uses platform-specific installation commands.

Implementation Details

Core Changes

• New types: PlatformOverrides and Variables structs in src/config.rs
• Validation for platform overrides in src/validate.rs
• Resolution logic in src/context.rs
• Compile-time platform selection using #[cfg(target_os = "...")]

Validation

• Only keys must be non-empty; values can be empty strings
• Supports clearing variables on specific platforms (e.g., tool_args = "")
• Reserved word "platform" is validated and rejected

Testing

• 186 tests pass on all platforms (Windows, Linux, macOS)
• 4 integration tests for platform-specific variables
• 7 unit tests for validation edge cases
• 98.12% line coverage in validate.rs

CI Improvements

• Windows coverage collection enabled (previously excluded)
• Platform-specific flags for Codecov uploads (Windows, Linux, macOS)
• Per-platform coverage analysis available on Codecov dashboard

Checklist

• All tests pass on all platforms
• CI checks pass (fmt, clippy, check, tests for Windows/Linux/macOS)
• Feature demonstrated in pls.toml
• Backward compatible
• Comprehensive test coverage

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 92.46575% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.80%. Comparing base (6406395) to head (59dd201).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
src/config.rs	86.84%	0 Missing and 5 partials ⚠️
src/context.rs	73.33%	0 Missing and 4 partials ⚠️
src/validate.rs	97.84%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #17      +/-   ##
==========================================
- Coverage   79.13%   78.80%   -0.33%     
==========================================
  Files          38       39       +1     
  Lines        6359     6742     +383     
  Branches     6359     6742     +383     
==========================================
+ Hits         5032     5313     +281     
- Misses        977     1069      +92     
- Partials      350      360      +10     

Flag	Coverage Δ
Linux	79.47% <91.66%> (?)
Windows	72.30% <91.66%> (?)
macOS	71.27% <92.36%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[claude]

Code Review

This PR adds Windows support and platform-specific variable overrides. The implementation is solid with good test coverage. Here are my findings:

Critical Issues

Platform override bug in Variables::resolve() (src/config.rs:36-55)
The cfg-based platform selection will fail on non-supported platforms (e.g., FreeBSD, Android). When none of the #[cfg(...)] match, platform_values is never defined, causing a compile error.

Recommendation: Add a catch-all case or explicitly handle unsupported platforms:

#[cfg(not(any(target_os = "windows", target_os = "linux", target_os = "macos")))]
let platform_values = &None;

Code Quality Issues

Inconsistent error handling for reserved word (src/validate.rs:46-51)
The reserved word "platform" check exists in validation but the error message suggests it's a deserialization error. The test at tests/test_variables.rs:107 expects a serde error, not a validation error, indicating the validation isn't actually preventing the issue.

The problem is that #[serde(flatten)] on values allows "platform" as a key, and serde parses it as a value before validation runs. You'd need #[serde(deny_unknown_fields)] and explicit field names to prevent "platform" being accepted as a variable.

Cloning overhead (multiple locations)

• src/config.rs:37,50 - Unnecessary clone() calls in resolve()
• src/config.rs:104-137 - Multiple clones in resolve_target_names_in()

Consider using references or into_iter() where ownership can be transferred.

Test Coverage Concerns

The codecov report shows 70% patch coverage with 24 lines missing. Key gaps:

• Error paths in Variables::resolve()
• Platform override edge cases in target_info_from_config()
• Some validation branches in validate_variables_option()

Minor Issues

• src/config.rs:40-41: Comment about "compile error" is misleading given the actual bug
• tests/test_variables.rs:107: Test name test_platform_reserved_word_error suggests validation error but checks for serde error
• Validation in src/validate.rs:55-64 could be simplified with a helper function or iterator

Positive Notes

• Clean separation of concerns with the Variables struct
• Good documentation in PR description and code comments
• Comprehensive test coverage for the happy path
• Backward compatibility maintained

Overall: Fix the platform selection bug before merging. Consider addressing the reserved word validation inconsistency.