fix(config): clamp max_chars to a minimum of 1

[YashManek1]

Problem

TRACEROOT_PLUGIN_MAX_CHARS set to 0 or a negative value silently disables span-text truncation instead of erroring or falling back — oversized span attributes can then hit backend/OTLP size limits and get dropped, and negative values produce corrupted, overlapping text.

Root cause

load_config() only guards against non-integer input (except ValueError), not out-of-range integers — int("0") and int("-5") both succeed. Then in middle_truncate, max_chars=0 makes half=0, and s[-half:] evaluates to s[-0:], which in Python returns the entire string. So the "truncated" output is actually the full text with a misleading "middle elided" marker glued on. Negative max_chars produces overlapping, corrupted slices.

Change

Defense in depth, two layers:

• config.py: clamp max_chars < 1 to the 20000 default, same as the non-integer fallback.
• spans.py: guard middle_truncate against half <= 0 directly, so any bad value that reaches it still degrades safely (elision marker only, no full-string leak).

Validation

• Added 5 regression tests in tests/test_spans.py: zero budget, negative budget, small positive budget (normal path), and two load_config clamp tests.
• Verified by hand: max_chars=0 -> half=0 -> s[-0:] == full string (confirms the bug); max_chars=-5 -> half=-3 -> s[3:] (confirms overlap).
• uv run pytest tests/test_spans.py -q -> 26 passed, 0 failed.
• Confirmed the new tests fail without the fix and pass with it.

Risk / rollback

Additive-only change to a config-bounds check and a defensive guard; no behavior changes on the default/valid-range path. Revert is a straight git revert of this commit.

Closes #8

---

Summary by cubic

Clamp invalid TRACEROOT_PLUGIN_MAX_CHARS values and harden middle_truncate so truncation can’t be disabled or corrupt span text. This keeps span attributes within size limits and prevents OTLP drops.

• Bug Fixes
  • Clamp max_chars < 1 to 20000 in load_config.
  • In middle_truncate, return only the elision marker when the budget is too small.
  • Added regression tests for zero/negative budgets and config clamping.

^{Written for commit e829633. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}