SG-44721 Fix tank core upgrade fails from ≤v0.23.6 to ≥v0.23.7

[julien-lang]

tank core fails with ModuleNotFoundError: No module named 'yaml'.

Two related cases:

1. Running _core_upgrader.py directly (e.g. python3 _core_upgrader.py) fails unconditionally on any version >= v0.23.7, including current master - there's no prior process state to mask it.
2. Running the real tank core upgrade command fails specifically when the currently installed core is <= v0.23.6 and the target is >= v0.23.7. Upgrading between two cores that are both already >= v0.23.7 (or both <= v0.23.6) is unaffected - see Root Cause below for why.

Root cause

SG-37222 / #996 reworked tank_vendor third-party dependency handling starting in v0.23.7: vendored packages moved from flat files under python/tank_vendor/ into zip archives, loaded lazily through a sys.meta_path finder registered in tank_vendor/__init__.py.

_core_upgrader.py was never updated for this change. It still added python/tank_vendor directly to sys.path and did a bare import yaml, bypassing tank_vendor/__init__.py entirely, so the finder never registered and the flat yaml module it expected no longer existed on disk.

tank core runs as a single, uninterrupted Python process (the tank -> tank_cmd.sh -> exec $interpreter tank_cmd.py chain replaces the process in place, and the core swap is a plain import _core_upgrader within that process). The currently installed core's core_upgrade.py already does from tank_vendor import yaml before the swap. If that core is already >= v0.23.7, its vendor package's meta finder resolves this via __import__("yaml"), which as a side effect registers bare sys.modules["yaml"] - masking the target's broken bare import yaml entirely. If the installed core is <= v0.23.6, its pre-rework tank_vendor has no meta finder, so nothing gets cached, and the target's import yaml fails for real. This is why the bug only reproduces on the <=v0.23.6 -> >=v0.23.7 transition, even though the underlying defect is unconditional.

Fix

• Add python/ (the parent of tank_vendor) to sys.path instead of tank_vendor itself.
• Import via from tank_vendor import yaml so tank_vendor/__init__.py runs and registers its lazy-loading finder.
• Use sys.path.insert(0, ...) instead of append for consistency with other core scripts.

Testing

• Ran _core_upgrader.py directly; it no longer raises ModuleNotFoundError on import.
• Confirmed the old code path fails the same way against current master's tank_vendor structure, matching the reported traceback.
• QA reproduced a real tank core upgrade from v0.23.6 -> v0.24.0 failing without the fix and succeeding with it, and validated v0.23.7 -> v0.24.0 as well; dedicated core-upgrade regression coverage has been added.

[Copilot]

Pull request overview

Fixes a core-upgrade regression introduced after the tank_vendor dependency-loading overhaul (v0.23.7+), where _core_upgrader.py imported yaml in a way that bypassed tank_vendor/__init__.py, causing ModuleNotFoundError during tank core upgrades.

Changes:

• Adds the local python/ folder (not python/tank_vendor/) to sys.path at position 0 so the bundled core code is preferred over any site-packages.
• Switches from import yaml to from tank_vendor import yaml to ensure tank_vendor’s import hook is registered.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.11%. Comparing base (d4cd9de) to head (26bb764).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1127   +/-   ##
=======================================
  Coverage   80.11%   80.11%           
=======================================
  Files         203      203           
  Lines       19540    19540           
=======================================
  Hits        15654    15654           
  Misses       3886     3886           

Flag	Coverage Δ
Linux	79.54% <ø> (ø)
Python-3.10	79.91% <ø> (ø)
Python-3.11	79.81% <ø> (ø)
Python-3.13	79.83% <ø> (+<0.01%)	⬆️
Python-3.9	79.88% <ø> (ø)
Windows	79.58% <ø> (ø)
macOS	79.51% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.

[chenm1adsk]

LGTM!