safe-global / safe-client-gateway / 37766663204

90%

Build Type

push

github

Committed by web-flow

Commit Message

fix(safe): order the queue by last modification on the queue service (#3491)

* fix(safe): order the queue by last modification on the queue service

getTransactionQueueByModified previously delegated straight to
proposeTransaction/postConfirmation without clearing the transaction and
queue caches on the queue-service path, and sorted the queue-service
queue by a transaction's own `modified` date only, missing a newer
confirmation that the upstream service does not always bump `modified`
for. Introduce getLastModified to take confirmations into account, clear
both cache layers after proposing/confirming, and use
SAFE_QUEUE_SERVICE_MAX_LIMIT (distinct from the tx-service's
SAFE_TRANSACTION_SERVICE_MAX_LIMIT) to bound the local sort.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* chore(safe): drop the unrelated eager queue-cache clear

Clearing the transaction and queue caches on propose and add-confirmation was scope unrelated to the queue ordering this branch changes, and it is not a demonstrated fix. Events reach the gateway over the AMQP fanout exchange and no producer is distinguished there, so a PENDING_MULTISIG_TRANSACTION or NEW_CONFIRMATION published by the queue service already clears the queue layer through clearMultisigTransactions, which invalidates both layers while FF_SAFE_QUEUE_SERVICE is on.

The integration test that appeared to pin the gap neither mocks the queue service nor flips the flag, so it exercises the tx-service path and passed only because the test harness has no AMQP producer.

Whether the queue service publishes to that exchange is a producer-side question. If it does not, the staleness is real and belongs in its own PR.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

Coverage Stats

4601 of 5391 branches covered (85.35%)

Branch coverage included in aggregate %.

19 of 19 new or added lines in 3 files covered. (100.0%)

2 existing lines in 2 files now uncovered.

11681 of 12790 relevant lines covered (91.33%)

548.92 hits per line