ISSUE DESCRIPTION auth: spurious REFRESHGENERATIONS bump on early-cancel in…
- User
ISSUE DESCRIPTION
auth: spurious _REFRESH_GENERATIONS bump on early-cancel in _coalesced_run_refresh_cmd
Summary
_coalesced_run_refresh_cmdinsrc/notebooklm/_auth/refresh.py(moved fromsrc/notebooklm/auth.pyin tier-10 PR-B-high #814, but the code is byte-identical to origin/main since tier-9 E #803) can bump_REFRESH_GENERATIONS[refresh_key]even when no real refresh happened, in a narrow cancel-before-await race.Reproducer (race condition)
- Caller invokes
await _coalesced_run_refresh_cmd(refresh_key, path, profile)from the cancel-recovery loop in_fetch_tokens_with_refresh. - Task is cancelled in the window between
try:and the firstwith _REFRESH_STATE_LOCK:block inside_coalesced_run_refresh_cmd— so NO future is registered in_REFRESH_INFLIGHT_BY_LOOP[loop][refresh_key]. - Control jumps to
except asyncio.CancelledError:in the caller. registry.get(refresh_key)returnsNone.- The inner
if inflight is None or inflight.done():is taken; theif inflight is not None:block is skipped;subprocess_excstaysNone;break. - Outer
if subprocess_exc is not None:is False, so no raise. - Generation is bumped at the
with _REFRESH_STATE_LOCK: ... max(existing, refresh_generation + 1)line, even though no subprocess ran. - Then
if caller_cancelled: raise asyncio.CancelledError().
Impact
A concurrent waiter on a different event loop sharing the same
refresh_keywould observe the phantom generation bump on its lock-re-acquire path, treat the storage as freshly-refreshed, and skip its own_run_refresh_cmd— proceeding tobuild_httpx_cookies_from_storageagainst STALE on-disk state.Bounded blast radius: the stale-storage caller will fail again on the next token fetch and trigger its own real refresh. No data corruption. But it's a wasted round-trip.
Suggested fix (from CodeRabbit on PR #814)
Track whether
inflightwas ever non-None during the cancel-recovery loop. Only bump generation ifsubprocess_exc is None AND observed_inflight is True. Or per the alternative suggestion: raise immediately wheninflight is Noneat the top of the cancel handler (the subprocess never registered, so we can't be the leader-cancel-before-await case for an actually-running refresh).Why not fixed in PR #814
That PR's scope was "extract verbatim from auth.py into _auth/refresh.py + facade-forward". Changing async semantics was explicitly out of scope (
must_not_do). Filing here for a dedicated follow-up.References
- CodeRabbit comment: https://github.com/teng-lin/notebooklm-py/pull/814#discussion_r3256142201
- Original code:
src/notebooklm/_auth/refresh.py:487-537(tier-10),src/notebooklm/auth.py:961-1026(tier-9 E origin)
Interface
No new interfaces are introduced.
ADDITIONAL ADVICE
Since you are given a git repository, you can use git commands to simplify your work. For example, if you made a mistake and want to revert your changes, you can use
git checkout <file>to restore the file to its original state. You can also reset all changes in the repository usinggit reset --hardcommand. Additionally, you can usegit stashandgit stash popto temporarily save your changes and restore them later. Finally, you can see the changes that you've made (which will be submitted when you callsubmit) usinggit statusorgit diffcommands. However you don't need to usegit addorgit commitbefore submitting your changes. In fact,submitonly submits currently unstaged changes because it usesgit diffto get the changes that need to be submitted.CHECKLIST
Before submitting your solution, please go over the following checklist and make sure you've done everything:
- If an error was reported in the issue description, I have successfully reproduced it.
- If an error was reported in the issue description, I have confirmed that my fix resolves the error.
- I have thought about edge cases and made sure my fix handles them as well.
- I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. I want you to list every bullet from this checklist and write your reasoning for why you think you did it or didn't need to.
Repository has been uploaded and your shell is currently at the repository root. Time to solve the issue!
(Current directory: /notebooklm-py, current file: none) bash-$
- Caller invokes
Assistant
Let me start by exploring the repository structure to understand what we're working with.
Assistant
Now let me look at
_coalesced_run_refresh_cmdfunction:Assistant
Now I have a clear picture. Let me also look at the existing tests for this code path:
Assistant
Now I have a thorough understanding. Let me look at the exact area around lines 487-537 of the file:
Assistant
Now I fully understand the issue. Let me also look at git history for context:
Assistant
Let me first create a reproducer to confirm the bug exists:
Assistant
The reproducer confirms the bug. Generation went from 0 to 1 even though no subprocess ever ran.
Now let me apply the fix. I'll go with the simpler "raise immediately when inflight is None" approach:
Assistant
Now I'll apply the fix:
Assistant
me use the
editcommand directly instead:Assistant
Now let me verify the fix works by running the reproducer:
Assistant
The fix works. Now let me think about edge cases and add more comprehensive tests.
Let me think through all the scenarios:
- Normal happy path: subprocess runs successfully → no cancellation → no exception → generation bumped ✓
- Subprocess fails: subprocess raises → except BaseException → subprocess_exc set → generation not bumped ✓
- Caller cancelled before inflight registered (the bug scenario): inflight is None → raise immediately → generation not bumped ✓ (new behavior)
- Caller cancelled while subprocess in flight (not yet settled): re-await shielded future, subprocess settles, success → break, bump generation, then raise CancelledError ✓
- Caller cancelled after subprocess settles with failure: inflight.done() → subprocess_exc set → generation not bumped ✓
- Caller cancelled after subprocess settles with success: inflight.done() → subprocess_exc is None → break → bump generation → raise CancelledError ✓
- Cancel/settle race: subprocess settles right when caller is cancelled → inflight.done() True → subprocess_exc set appropriately → generation not bumped on failure ✓ (already covered)
Let me also make sure the existing tests still pass. Let me also add a test scenario for case 4 specifically — caller cancelled while subprocess is in-flight.
Let me first run the existing test suite for the refresh tests:
Assistant
All existing tests pass. Let me also run the broader auth tests:
Assistant
All auth and refresh related tests pass. Let me also run the broader unit tests:
Assistant
The failures look like they're related to playwright/browser testing, not to my changes. Let me verify by checking against the unmodified state:
Assistant
: the playwright tests fail without my changes due to missing playwright module. These are pre-existing environmental failures, not regressions from my fix.
Let me run the rest of the test suite excluding playwright tests:
Assistant
All 4534 unit tests pass. Let me also run the integration tests, excluding the playwright-dependent ones:
Assistant
All 670 integration tests pass. Now let me extend my reproducer to cover additional edge cases:
Assistant
All edge cases pass. Now let me also verify the fix works for the case where the future IS in the registry (cancel during in-flight subprocess) — the existing tests already cover this, but let me also write a reproducer for case 4 and case 5 to be thorough:
Assistant
All edge cases pass correctly. Let me also verify the fix works through the
auth.pyfacade (since the issue mentions the facade write-through):Assistant
The fix works correctly through the facade as well. Let me also do a thorough re-review of the diff to make sure it's correct:
Assistant
The diff is minimal and focused. Let me also clean up the unused
if inflight is not Nonecheck inside theif inflight.done():block (since we now knowinflightis non-None at that point):Assistant
Let me also verify the inner
if inflight is not None:block is still correct (now unreachable but harmless due to the early raise). The defensive check helps with mypy narrowing. I'll leave it as-is to minimize the change.Let me run the relevant tests one more time:
Assistant
All tests pass. Let me also run mypy to make sure the type annotations are still correct:
Assistant
No mypy errors. Let me also run ruff to make sure formatting/linting is clean:
Assistant
All checks pass. Let me also look at the final state of the fix:
Assistant
Let me verify the bug is present without the fix by temporarily reverting and running the reproducer:
Assistant
The fix is verified:
- Without the fix: phantom generation bump (gen=1) — bug present
- With the fix: no phantom generation bump (gen=0) — bug fixed
Let me clean up my reproducer scripts and ensure the fix is final:
Assistant
The fix is complete. Let me run one more comprehensive test to make sure nothing is broken:
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.5-35B-A3BCompare side by side →
Qwen/Qwen3.5-35B-A3BCompare side by side →