Match the whole value in env2bool

[shashvat-singham]

Fixes #11080

env2bool decided truthiness with an unanchored re.search, so any value that merely contained 1, y, yes or true was treated as true:

DVC_EXP_AUTO_PUSH=my_path   -> True    # contains "y"
DVC_EXP_AUTO_PUSH=anything  -> True    # contains "y"
DVC_EXP_AUTO_PUSH=v1.0      -> True    # contains "1"
DVC_EXP_AUTO_PUSH=nay       -> True    # contains "y"

It was also asymmetric in a way that's hard to reason about: no is correctly false, but nay is true.

This matters because these are user-facing environment variables where the input is arbitrary text — DVC_EXP_AUTO_PUSH, DVC_SQLALCHEMY_ECHO, DVC_IGNORE_ISATTY, DVC_TEST. The likely way to hit it is a value meant to be negative that happens to contain a matching letter, or a non-boolean value put in one of these by mistake, which silently switches the feature on instead of being ignored.

Change

Compare the whole value, after strip() and lower(), against the set of accepted affirmatives.

Every value that was intentionally supported still works (1, y, yes, true, and case/whitespace variants); only substring matches stop being truthy. The undefined= behaviour for an unset variable is unchanged.

Note this is technically a behaviour change for anyone currently relying on a loose value being truthy — but since that path is indistinguishable from "user set something that isn't a boolean", tightening it seems clearly right. Happy to add a warning on unrecognised values instead if you'd prefer a softer landing.

Tests

test_env2bool is parametrised over the accepted affirmatives, the common negatives, and the five substring cases above; test_env2bool_undefined pins the unset behaviour.

$ pytest tests/unit/utils/test_utils.py -k env2bool
17 passed

Five of them fail on main.

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.98%. Comparing base (2431ec6) to head (b4499a1).
⚠️ Report is 213 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #11082      +/-   ##
==========================================
+ Coverage   90.68%   90.98%   +0.30%     
==========================================
  Files         504      505       +1     
  Lines       39795    41146    +1351     
  Branches     3141     3263     +122     
==========================================
+ Hits        36087    37436    +1349     
- Misses       3042     3071      +29     
+ Partials      666      639      -27     

☔ View full report in Codecov by Harness.
📢 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.

[shashvat-singham]

On the red lint checks — they're pre-existing on main, not from this PR.

The failing pre-commit hook is mypy, and the error is in a file this PR doesn't touch:

dvc/commands/completion.py:20:63: error: Argument "preamble" to "complete" has ...

The same hook fails with the same error on main's latest run (job 95098111969), across all three OSes.

Everything else in pre-commit passes on the changed files here — ruff check, ruff format, codespell, end-of-file/line-ending/whitespace hooks. Happy to rebase once main is green if you'd like a clean run.

[shashvat-singham]

Note on the red lint jobs: they are not from this change.

pre-commit fails at the mypy hook on dvc/commands/completion.py:20 — Argument "preamble" to "complete" has incompatible type "dict[str, str]"; expected "str" — a file this PR does not touch. The same three lint jobs (ubuntu/macos/windows) fail identically on main at the base commit, so it is pre-existing rather than something the branch introduced.

Everything else in the run, including ruff check, ruff format and the env2bool tests, passes.