DotNetWorkQueue v0.12.0

What's Changed

• Move the five remaining actions off Node 20 by @blehnen in #267
• Build each stage's own test project instead of the whole solution by @blehnen in #269
• Run integration tests in parallel within each assembly by @blehnen in #270
• Gate on the compile, run unit tests beside the integration stages by @blehnen in #272
• Stop the SQLite retry policy tests sitting through the whole retry ladder by @blehnen in #273
• Stagger longest stage first, and drop the retry flag that never ran by @blehnen in #274
• Restore from an internal NuGet mirror when the controller offers one by @blehnen in #275
• Scope the GITHUB_TOKEN in ci.yml and sonarcloud.yml by @blehnen in #276
• Design: async receive path (#256) by @blehnen in #277
• Plan: PR 1 of the async receive path — the awaitable throttle by @blehnen in #278
• Awaitable throttle for the consumer backpressure primitive (#256, PR 1 of 7) by @blehnen in #279
• Plan: PR 2 of the async receive path — IReceiveMessages and the gate by @blehnen in #280
• Async receive path: core plumbing plus a genuinely async Redis de-queue (#256) by @blehnen in #282
• An async command-handler abstraction for the no-output commands (#284) by @blehnen in #285
• Move the relational ADO abstraction onto System.Data.Common (#286) by @blehnen in #287
• feat: move a poison message to the error queue without blocking (#284) by @blehnen in #289
• feat: an awaitable remove, and the async delete handlers under it (#284) by @blehnen in #291
• feat: record the processing-start history row without blocking (#284) by @blehnen in #293
• feat: commit a message without blocking on the async consumer (#284) by @blehnen in #295
• feat: roll a message back without blocking on the async consumer (#284) by @blehnen in #296
• Handle a failed message without blocking the async consumer (#284) by @blehnen in #297
• A heartbeat that neither blocks the async consumer nor holds a pool thread (#284) by @blehnen in #301
• A worker whose heartbeat has lapsed no longer keeps the message (#294) by @blehnen in #302
• Three remaining correctness and blocking issues (#288, #290, #292) by @blehnen in #305
• Make the move to the error queue atomic under a held transaction (#300) by @blehnen in #306
• Count message errors in one statement, where the schema allows it (#299) by @blehnen in #307
• History timestamps come from the configured time provider (#304) by @blehnen in #310
• The rest of the timestamps that bypassed the time provider (#304) by @blehnen in #312
• Drive heartbeats from a timer instead of the job scheduler (#303) by @blehnen in #315
• History timestamps read back as the UTC they were written as (#311) by @blehnen in #317
• Serialise LiteDb table creation (#318) by @blehnen in #319
• Document why retries cannot succeed inside a caller-held transaction (#309) by @blehnen in #322
• Document SQLite's synchronous async path as a permanent constraint (#298) by @blehnen in #323
• SQLite: set WAL journal mode on a database a job producer creates by @blehnen in #325
• Test that the stale cut-off and Memory's QueuedDateTime come from the configured clock (#313) by @blehnen in #326
• Stop the metric verifiers pretending they can wait for a count to land (#314) by @blehnen in #327
• Publish to NuGet with trusted publishing instead of a long-lived API key (#266) by @blehnen in #330
• Give SQLite's heartbeat test a claim window its writes can actually meet (#328) by @blehnen in #329
• Bound the Redis wait for work, so a lost notification is not permanent (#331) by @blehnen in #332
• Restore the metric verifier polling removed in #327 by @blehnen in #333
• Clear the code smells accumulated since 0.11.0 (#324) by @blehnen in #334
• Stop losing messages to a late heartbeat, and prove the claim before refreshing it (#328) by @blehnen in #335
• Clear the code smells the #328 heartbeat work left behind by @blehnen in #337
• Publish a tested upgrade script for 0.12.0 by @blehnen in #338
• Release 0.12.0 by @blehnen in #340

Full Changelog: v0.11.0...v0.12.0