fix(env): parse --env=VALUE and -E=VALUE flag forms correctly

[jdx]

Summary

• Fix --env=production and -E=production (equals-sign form) being silently ignored, causing fallback to the default environment from .miserc.toml
• The environment() function in src/env.rs parsed args using windows(2) which only matched space-separated flag/value pairs (--env production), not the =-joined form
• Now handles both --env production and --env=production (and likewise -E, -P, --profile)

Fixes #8883

Test plan

• mise --env=production env produces same output as mise --env production env
• mise -E=production env produces same output as mise -E production env
• mise --env production env still works as before
• mise -E production env still works as before
• Comma-separated values work: mise --env=dev,staging env

🤖 Generated with Claude Code

---

Note

Medium Risk
Adjusts CLI argument parsing for environment selection, which can change which config/environment is applied at runtime; scope is localized and covered by new e2e tests.

Overview
Fixes environment/profile flag parsing so --env=dev, -E=dev, --profile=dev, and -P=dev are recognized (in addition to space-separated forms), including comma-separated multi-env values.

Updates environment() scanning to stop treating flags after mise run <task> as global env selectors, and adds an e2e test validating equals-sign, space-separated, comma-separated, and mise run ordering behavior.

^{Reviewed by Cursor Bugbot for commit 5c3df3b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[gemini-code-assist]

Code Review

This pull request updates the argument parsing logic in src/env.rs to support both space-separated (e.g., --env production) and equals-separated (e.g., --env=production) flags. A logic flaw was identified where space-separated values are not skipped, potentially leading to double-processing of arguments. A more efficient implementation using an iterator and split_once('=') was suggested to resolve this and simplify the code.

[gemini-code-assist]

The current implementation of the argument parsing loop has a logic flaw: it does not skip the next argument when it is consumed as a value for a space-separated flag (e.g., --env production). This causes the value to be processed again as a potential flag in the next iteration. For example, with mise --env --env=prod, the current code would add both --env=prod and prod to the values list.

Additionally, the logic can be simplified and made more efficient by using an iterator and split_once('='), which avoids the need to pre-allocate arg_prefixes and the intermediate args_before_dashdash vector.

Suggested change

	let arg_prefixes: Vec<String> = arg_defs.iter().map(|a| format!("{a}=")).collect();
	// Try to get from command line args first
	args.windows(2)
	.take_while(|window| !window.iter().any(|a| a == "--"))
	.filter_map(|window| {
	if arg_defs.contains(&*window[0]) {
	Some(window[1].clone())
	} else {
	None
	// Handles both `--env production` (separate args) and `--env=production` (joined with =)
	let mut values = Vec::new();
	let args_before_dashdash: Vec<_> = args.iter().take_while(|a| a.as_str() != "--").collect();
	for (i, arg) in args_before_dashdash.iter().enumerate() {
	if let Some(prefix) = arg_prefixes.iter().find(|p| arg.starts_with(p.as_str())) {
	values.push(arg[prefix.len()..].to_string());
	} else if arg_defs.contains(arg.as_str()) {
	if let Some(next) = args_before_dashdash.get(i + 1) {
	values.push(next.to_string());
	}
	})
	}
	}
	let mut values = Vec::new();
	let mut it = args.iter().take_while(|a| a.as_str() != "--");
	while let Some(arg) = it.next() {
	if let Some((flag, value)) = arg.split_once('=') {
	if arg_defs.contains(flag) {
	values.push(value.to_string());
	}
	} else if arg_defs.contains(arg.as_str()) {
	if let Some(next) = it.next() {
	values.push(next.to_string());
	}
	}
	}
	values

[greptile-apps]

Greptile Summary

This PR fixes a silent bug where --env=VALUE and -E=VALUE (equals-sign forms) were ignored by environment() in src/env.rs, causing fallback to the default environment. The root cause was that windows(2) only matched space-separated flag/value pairs.

The fix replaces the windows(2) approach with an explicit iterator loop that handles both forms, and adds an in_run_subcommand guard so that -E/--env appearing after the task name in mise run <task> is correctly treated as a task argument rather than a global flag.

Key changes:

• src/env.rs — new iterator-based arg scanner handles --env=VALUE, -E=VALUE, --profile=VALUE, -P=VALUE, comma-separated values, and the mise run positional-arg boundary correctly.
• e2e/cli/test_env_flag_forms — comprehensive new test covering the fixed forms, regression cases for space-separated forms, and mise run interaction — however line 53 contains a typo (mise run --env=dev run show-env has a spurious extra run that will cause the test to fail since no task named run is defined in the fixture).

Confidence Score: 4/5

Safe to merge after fixing the typo on line 53 of the e2e test — the core env.rs logic is correct and well-reasoned.

The src/env.rs implementation is sound: it correctly handles all four flag forms, comma-separated values, and the mise run positional-arg boundary. One P1 defect exists in the new e2e test — line 53 uses mise run --env=dev run show-env (with a spurious extra run), which will fail at runtime because the task named run does not exist in the fixture. This test would either fail CI or pass only by accident; it needs a one-word fix before merging.

e2e/cli/test_env_flag_forms line 53 — spurious extra run in the command string causes the test to attempt running a non-existent task.

Important Files Changed

Filename	Overview
src/env.rs	Replaces windows(2)-based arg scanning with an explicit iterator loop that handles both space-separated (--env prod) and equals-joined (--env=prod) forms; adds in_run_subcommand guard to stop scanning after the task name in mise run.
e2e/cli/test_env_flag_forms	New e2e test covering equals-sign, space-separated, comma-separated, and run-subcommand interaction forms — but line 53 has mise run --env=dev run show-env (duplicate run) which will fail because no task named run exists in the test fixture.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["environment(args)"] --> B{IS_RUNNING_AS_SHIM?}
    B -- Yes --> C[return vec! empty]
    B -- No --> D[iter over args, stop at --]
    D --> E{next arg?}
    E -- None --> F[collect values]
    E -- Some arg --> G{arg contains '='?}
    G -- Yes --> H{flag part in arg_defs?}
    H -- Yes --> I[push value part] --> E
    H -- No --> E
    G -- No --> J{arg in arg_defs?}
    J -- Yes --> K[it.next as next_arg
push next_arg] --> E
    J -- No --> L{arg starts with '-'?}
    L -- Yes --> E
    L -- No --> M{in_run_subcommand?}
    M -- Yes --> N[BREAK]
    M -- No --> O{arg in run_subcommands?}
    O -- Yes --> P[in_run_subcommand = true] --> E
    O -- No --> E
    F --> Q[flat_map: split each value on comma]
    Q --> R{from_args empty?}
    R -- No --> S[return from_args]
    R -- Yes --> T[check MISE_ENV / MISE_PROFILE / .miserc.toml]

Loading

_{Reviews (4): Last reviewed commit: "refactor(env): simplify flag parsing wit..." | Re-trigger Greptile}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2cb67b9. Configure here.}

[cursor]

False run-subcommand detection from other flags' values

Low Severity

The in_run_subcommand heuristic doesn't account for values of other global flags being misidentified as positional args. Since environment() only knows about --env/-E/--profile/-P, the values of other value-taking flags like --cd/-C fall through to the positional-arg branch. If that value happens to be "run" or "r" (e.g., mise -C run env -E dev), in_run_subcommand is falsely set to true, causing subsequent -E flags to be silently ignored. The old windows(2) code didn't have this issue since it had no in_run_subcommand guard. Having a directory named run is plausible in real projects.

 

^{Reviewed by Cursor Bugbot for commit 2cb67b9. Configure here.}

[github-actions]

Hyperfine Performance

xtasks/test/perf

Command	mise-2026.4.3	mise	Variance
install (cached)	116ms	116ms	+0%
ls (cached)	66ms	66ms	+0%
bin-paths (cached)	68ms	68ms	+0%
task-ls (cached)	719ms	⚠️ 2636ms	-72%

⚠️ Warning: task-ls cached performance variance is -72%