- Added rescue block around backfill_registry so failures are
  best-effort (logged + telemetry) rather than propagating
- Replaced all skip-based tests with mock-based tests that exercise
  backfill logic without requiring the compiled C extension
- Added tests for: mixed iseq types, multiple files, error handling,
  suffix/exact lookup on backfilled entries, start ordering
- 27 examples, 0 failures, 0 pending, 0 skipped

Co-Authored-By: Claude <[email protected]>

Tests the end-to-end flow: test class loaded before code tracking
starts → CodeTracker#start triggers backfill via all_iseqs C extension
→ iseq recovered from object space → line probe installed on
backfilled iseq → probe fires and captures local variables.

Runs under rake spec:di_with_ext (requires compiled C extension).

Three test cases:
- Probe installs successfully on backfilled iseq
- Probe fires when target line executes
- Snapshot captures local variables from backfilled iseq

Co-Authored-By: Claude <[email protected]>

On macOS CI the C extension is compiled, so backfill_registry populates
the CodeTracker registry with pre-loaded files during start. This broke
existing tests that expect the registry to be empty after start or to
contain exactly N explicitly-loaded files.

Fix by stubbing backfill_registry in test contexts that exercise
:script_compiled behavior. Backfill is tested separately in its own
describe blocks.

Affected contexts:
- CodeTracker #start (before block)
- CodeTracker shared context 'when code tracker is running'
- CodeTracker #iseqs_for_path_suffix (around block)
- Instrumenter shared context 'with code tracking'

Co-Authored-By: Claude <[email protected]>

…kfill

The backfill filter used first_lineno == 0 to identify whole-file iseqs,
but most whole-file iseqs from all_iseqs have first_lineno == 1. The new
DI.iseq_type method reads the iseq type directly from the Ruby VM struct
and returns a symbol (:top, :method, :block, :class, etc.).

The backfill now filters by type == :top || type == :main, which
correctly identifies whole-file iseqs regardless of first_lineno.

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

rb_iseq_type is an internal Ruby function that only exists in Ruby 3.1+.
On Ruby 2.7 and 3.0, referencing it causes an undefined symbol error at
load time, crashing the entire C extension (including all_iseqs and
exception_message which work fine on those versions).

Use have_func in extconf.rb to detect rb_iseq_type at compile time, and
wrap the iseq_type function + registration in #ifdef HAVE_RB_ISEQ_TYPE.
The Ruby code in code_tracker.rb already handles the missing method via
DI.respond_to?(:iseq_type) with a first_lineno fallback.

Co-Authored-By: Claude <[email protected]>

- Remove load_iseq from prepended module in after block instead of
  leaving the prepend as a permanent global mutation. Uses the same
  pattern as the DI instrumenter for method probe cleanup: the module
  stays in the ancestor chain but becomes a no-op.
- Regenerate ruby_3.2_rails8_mysql2.gemfile to include bootsnap.
  Other Ruby version gemfiles will be regenerated in their respective
  CI environments.

Co-Authored-By: Claude <[email protected]>

- Rewrite `unless` with mixed logical operators to `if` with negated
  conditions in backfill_registry (Style/UnlessLogicalOperators)
- Remove redundant begin block and line continuations in Bootsnap test

Co-Authored-By: Claude <[email protected]>

…oads

The skip guard only caught LoadError, but bootsnap can load successfully
while its iseq cache fails to initialize (e.g. C extension not compiled
for this environment). This caused test failures in CI jobs where
bootsnap was available as a transitive dependency but couldn't install
its iseq cache.

Fix: probe-test the iseq cache in before(:all) — install it in a temp
dir, clean up, and skip if any error occurs.

Co-Authored-By: Claude <[email protected]>

The require was inside before(:all), which loads bootsnap into the
process at test execution time — after other test infrastructure has
initialized. Moving it to file load time makes the side effect visible
and predictable. If bootsnap isn't available or its iseq cache can't
initialize, the constant BOOTSNAP_AVAILABLE is false and the entire
describe block skips via before(:all).

Also eliminates the standard:disable comment for Style/RedundantBegin
since the rescue is now in a top-level begin/rescue/end, not a block.

Co-Authored-By: Claude <[email protected]>

… RBS type

- compile_file_backfill_spec.rb: fix skip message "Ruby < 3.1" → "Ruby >= 3.1 only"
- iseq_type_spec.rb: rename "eval with top-level code" → "compiled top-level code"
  (the test uses compile(), not eval)
- code_tracker_spec.rb: remove two comments that just restate the expectation below them
- code_tracker_spec.rb: rename "calls backfill_registry after trace point is enabled"
  test — it didn't actually verify ordering, only that active? is true after start
- sig/datadog/di.rbs: iseq_type return type Symbol → Symbol? (C code returns Qnil
  when rb_iseqw_to_iseq returns NULL)

Co-Authored-By: Claude Sonnet 4.6 (1M context) <[email protected]>

The guard fires in zero CI jobs — the di:bootsnap job always has bootsnap
in the gemfile. It only fires when a developer runs the file manually without
the correct gemfile, in which case a LoadError is the right signal.

More importantly, the StandardError rescue around install!() could silently
skip all tests if bootsnap's C extension fails on a new Ruby version, hiding
a real compatibility break as green CI.

Co-Authored-By: Claude Sonnet 4.6 (1M context) <[email protected]>

The test "does not register compile_file iseqs" passed vacuously when
the code was correct — the `if result` block meant zero assertions ran
when result was nil. Add `expect(result).to be_nil` to directly assert
the intended behavior.

Co-Authored-By: Claude Sonnet 4.6 (1M context) <[email protected]>

require "bootsnap" alone does not load Bootsnap::CompileCache::ISeq —
that submodule must be required explicitly. Without it, the spec fails
with NameError when referencing Bootsnap::CompileCache::ISeq.

Also restores the BOOTSNAP_AVAILABLE skip guard that was removed, so
the tests degrade gracefully when bootsnap or its ISeq cache is
unavailable.

Co-Authored-By: Claude Sonnet 4.6 (1M context) <[email protected]>

The di:bootsnap CI task runs with rails8-mysql2 gemfile but does not
compile the DI C extension (that is done by di_with_ext). When
DI::Component.build calls environment_supported?, it finds
DI.respond_to?(:exception_message) is false and calls logger.warn().
The strict instance_double rejects the unexpected warn call, failing
the test before it can exercise any Bootsnap behavior.

Add a skip guard parallel to the existing BOOTSNAP_AVAILABLE guard.

Co-Authored-By: Claude <[email protected]>

- Add @return [void] YARD tag to backfill_registry
- Restructure compile_file_backfill_spec into two contexts:
  - "with iseq_type (Ruby 3.1+)" — existing test, skips on < 3.1
  - "with first_lineno fallback (all Ruby versions)" — new test that
    stubs iseq_type away and verifies the first_lineno == 0 heuristic
    also excludes compile_file iseqs (which have first_lineno == 1)
- Extract shared setup into helper method to avoid duplication

Co-Authored-By: Claude <[email protected]>

The previous fix added a skip guard for the DI C extension, which is a
build dependency not a platform limitation. That guard silently skipped
all bootsnap tests on every CI run.

Correct fix: add the compile dependency to the di:bootsnap Rake task,
matching the pattern used by di:di_with_ext. Remove the skip guard.

Co-Authored-By: Claude <[email protected]>

The di:bootsnap CI job runs against rails8-mysql2 gemfiles which include
bootsnap. The skip guard fires in zero CI jobs — it only fires when a
developer runs the file manually without bootsnap installed. That's a
local dev convenience skip, not a platform limitation.

If bootsnap can't load, the test should fail — that tells the developer
they need the right gemfile. Requires and initialization now happen at
file load time without rescue, so failures are immediate and visible.

Co-Authored-By: Claude <[email protected]>

The file-level Bootsnap::CompileCache::ISeq.install!(BOOTSNAP_VERIFY_DIR)
set Bootsnap's internal cache directory, which was then deleted.
Subsequent install!(cache_dir) calls in the before block may not update
the internal state, causing cache files to be written to the deleted
directory. Other bootsnap tests passed because they don't check for
cache files on disk.

Fix: remove the file-level install!/cleanup entirely. The requires
at file level are sufficient to prove bootsnap loads. The test's
before block handles install! with the correct cache directory.

Co-Authored-By: Claude <[email protected]>

Co-Authored-By: Claude <[email protected]>

Bootsnap's load_iseq intentionally returns nil when Coverage is active
(line 53 of bootsnap/compile_cache/iseq.rb), disabling iseq disk
caching entirely. This means no cache files are written during CI runs
with SimpleCov, causing the "Bootsnap cache was actually used" test to
fail — it checks for cache files on disk.

This is a genuine platform condition: Bootsnap + Coverage are
incompatible for disk caching. Other bootsnap tests (iseq capture,
TracePoint firing, probe installation, snapshot capture) pass because
they test the :script_compiled path, not the disk cache.

Root cause of the previous incorrect fix (removing file-level install!):
the file-level install! was not the problem — Coverage was.

Co-Authored-By: Claude <[email protected]>

This reverts commit 45a9f83281924b2b4321982b66bf619835ce7abf.

Bootsnap's load_iseq intentionally returns nil when Coverage is running
(SimpleCov in CI), disabling iseq disk caching entirely. This caused
the "Bootsnap cache was actually used" test to fail — no cache files
were written during the priming load.

Fix: suspend Coverage for the duration of each bootsnap test example.
Coverage.suspend (Ruby 3.2+) pauses coverage collection, allowing
Bootsnap's load_iseq to proceed normally. di:bootsnap runs on 3.2+
only, so Coverage.suspend is always available.

The suspend/resume is tracked via @coverage_was_suspended to avoid
calling Coverage.resume in local dev where Coverage was never started
(which would raise RuntimeError).

Co-Authored-By: Claude <[email protected]>

Two bugs caused "Bootsnap cache was actually used" to always fail:

1. ISeq.install!(cache_dir) sets cache_dir to "#{cache_dir}-iseq"
   internally but does not create the directory. Bootsnap's native
   fetch silently skips writing when the dir doesn't exist. Fix:
   mkdir_p the actual iseq cache dir after install!.

2. Dir.glob checked cache_dir but Bootsnap writes to cache_dir + "-iseq".
   Fix: glob the actual iseq cache dir from ISeq.cache_dir.

Also fix the after block: remove_method :load_iseq from the prepended
mixin is permanent — prepend is idempotent, so a subsequent install!
won't re-add load_iseq, breaking later examples. Instead, point the
cache dir at the deleted directory so load_iseq becomes a no-op.

The Coverage.suspend added in the previous commit is still needed:
Bootsnap's load_iseq returns nil when Coverage.running? is true,
bypassing the cache path entirely.

Co-Authored-By: Claude <[email protected]>

Address review comment: vpellan suggested initializing to [] to
simplify the type and remove the steep:ignore workaround.

- @current_components = [] (was nil)
- Removed ||= [] lazy init in add_current_component
- Removed &. safe navigation (no longer nullable)
- Removed steep:ignore NoMethod
- Updated RBS: ::Array[Component] (was ::Array[Component]?)

Co-Authored-By: Claude <[email protected]>