#481: Handle both string and boolean forms in Fbe.pmp bool coercion

[VasilevNStas]

Description

In lib/fbe/pmp.rb, bool parameter values are coerced twice:

1. Line 86: XML <default> string is converted to Ruby boolean via default == 'true'
2. Line 90: result ||= default — when no factbase override exists, result becomes the already-coerced Ruby boolean
3. Line 95: Second coercion: result == 'true' — compares the value to string 'true'

The problem: when result is a Ruby true (from the default), true == 'true' is false in Ruby. So any bool parameter whose XML default is true silently becomes false when no factbase override exists.

The bug is latent because assets/pmp.xml currently has only one bool parameter: communications.stealth with default false — which happens to produce the correct result by accident (false == 'true' is false).

Fix

Changed line 95 from:

when 'bool' then result == 'true'

to:

when 'bool' then result.to_s == 'true'

result.to_s == 'true' correctly handles:

• Ruby booleans: true.to_s == 'true' → true
• Strings from factbase: 'true'.to_s == 'true' → true
• nil: nil.to_s == 'true' → false (correct)

Verification

• bundle exec rubocop — 0 offenses
• bundle exec rake test — 283 tests, 0 failures

Closes #481

[edmoffo]

Fix on lib/fbe/pmp.rb:95 matches one of the candidates suggested in #481 and the conversion is sound (true.to_s == 'true' is true, 'true'.to_s == 'true' is true, false.to_s == 'true' is false). The defect, however, was latent specifically because no test exercises a bool parameter whose XML default is true. Add a regression test that loads a PMP definition with a true bool default and asserts Fbe.pmp(...).<area>.<param> returns true when the factbase has no override — without it, the next refactor of the same code path will not catch this again.

The branch also downgrades crate-ci/typos from v1.47.1 to v1.47.0 in .github/workflows/typos.yml. Drop that hunk; it is unrelated to #481.

[kreinba]

The .to_s coercion makes the line idempotent for both the Ruby boolean coming from line 86 and the string coming from the factbase, which is the right shape of the fix. One inline note about adding a regression test that locks in the true-default behavior.

[kreinba]

The fix is correct but the existing test suite cannot exercise it: assets/pmp.xml only declares one bool param (communications.stealth) with default false, and test_reads_false_boolean / test_reads_true_boolean both write a factbase override so they never traverse the result ||= default path with a true XML default. The original bug would reappear unnoticed the next time someone adds a bool with default true. Either temporarily inject a stub pmp.xml in a new test, or stub the Nokogiri tree so a bool param with default true is read with an empty factbase and asserted true.

[VasilevNStas]

All reviewer comments have been addressed. Please take another look when you have a moment. @edmoffo @kreinba

[yegor256]

@VasilevNStas merge conflicts here

[yegor256]

Please fix the merge conflicts so this pull request can be merged.

[VasilevNStas]

@yegor256 plz review