Repository navigation
test: realign two main-side pins broken by merge races - #14083
Closed
iamwhatever wants to merge 1 commit into
Closed
iamwhatever wants to merge 1 commit into
iamwhatever wants to merge 1 commit into
Conversation
Two tests fail on main because each pair of PRs was green alone and merged against a base that lacked the other. test_sandbox_protected_name_holds: the leaf-only ratchet was measured before two new protected names reached main: the auth-store staging directory and the credential-redaction owner switch. Each is protected under all three data-home roots the launcher emits, so every tier grows by exactly six (standard 235->241, cc 242->248, strict 243->249). Neither name has a stand-in-masked ancestor that could hold it, so the pin moves to the measured count with the reason in the comment. No protected name was removed and no hold regressed. test_spawn_approval_channel_delivery_2381: the silent-send test read any registry entry during the send as "a wait". The window is claimed at arm time, before the send, by design, so that entry is the claim, not a wait. The probe now counts the decider being awaited, and also asserts the claim is fully released (registry, nonce and ownership) after the fall-through. Disabling the fall-through still fails the test.
Contributor
Author
|
Superseded by #14056, which landed the same two fixes (pin 241/248/249 with the same two names; spawn probe counts the decider wait). Both files pass on main 64/64. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Two tests are red on
main(be067a3503) and on every PR branched from it:Backend Tests (3.12, 6),(3.12, 8),Backend Tests (Windows) (6), andCoverage Gateas the cascade. Both are merge races: each pair of PRs was green alone, and each merged against a base without the other. Both fixes are test-side; no product code was wrong.test_sandbox_protected_name_holds.py::TestLeafOnlyPopulationIsRecorded::test_leaf_only_count_has_not_grown(3 tiers)auth-store-staging) + #13685 (credential_redaction.json)test_spawn_approval_channel_delivery_2381.py::...::test_it_registers_no_wait_so_the_window_is_never_spent19fca6422cnever contained #13609, which reserves the window at arm time, before the send, by design1. Protected-name leaf-only pin (security pin, not weakened)
Diffing the leaf-only set at #13935's merge
2a4457f9f2againstbe067a3503, inside the pytest env: only additions, no removals, exactly six per tier:<root>/auth-store-staging(fix(sandbox): refuse a masked credential leaf that has a second hard link #13281: the masked staging dir both auth-store publishers stage in)<root>/credential_redaction.json(feat(security): owner switch to turn credential redaction off in the file viewer #13685: the owner switch for file-viewer redaction; authority that must not be agent-writable)each under the three data-home roots the launcher emits. Neither has a stand-in-masked ancestor, so neither can be held by an enclosing mask without restructuring the data home.
EXPECTEDbecomesstandard 241 / cc 248 / strict 249, with the reason in the comment. The hold-regression tests in the same file are unchanged and pass.Note: #12684 (open feature PR) also edits this pin to include its own additions; whichever lands second rebases the number.
2. Spawn-approval silent-send probe (stale after an intended change)
The test's probe read
key in TelegramApprovalDecider._REGISTRYduring the send as "a wait was registered". Since #13609,arm(..., detached=True)reserves the future and claims ownership before the send on purpose (so a press in that gap is not lost and the end-of-turn sweep leaves it alone). The #13799 fall-through itself is correct:retire()releases everything.The probe now wraps
TelegramApprovalDecider.__call__and asserts it is never awaited, plus asserts the key is gone from_REGISTRY,_NONCESand_AWAITEDafter the fall-through. Mutation check: replacingif not posted:withif False:intransport_dispatch.deliver_spawn_approvalfails the test.Testing
pytest test/test_sandbox_protected_name_holds.py test/test_spawn_approval_channel_delivery_2381.py: 64 passed (4 failed before).origin/mainbe067a3503; the spawn test also fails at fix(telegram): fall through when the spawn-approval prompt is not accepted #13799's own merge commit7f1f4fb1d7.36208417322: Windows (6) and 3.12 (6) fail on the spawn test; 3.12 (8) on the three pin cases. Same tests, no third failure.black --check,check_comment_history.py(baseorigin/main) pass.