Merge pull request #69 from Sendspin/fix/release-0-3-0

Release 0 3 0

Author: chrisuthe

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.16)
 
 project(sendspin-cpp-cli
-    VERSION 0.2.0
+    VERSION 0.3.0
     DESCRIPTION "Headless CLI/daemon Sendspin audio player"
     HOMEPAGE_URL "https://github.com/Sendspin/sendspin-cpp"
     LANGUAGES CXX

docs/ROADMAP.md

@@ -2639,7 +2639,7 @@ The harness it was run with is not in the tree: it lives with the pass's own not
 sink suite that exercises `AudioSink` implementations themselves is item 12's job, not a
 throwaway driver's.
 
-### 28. A second device outage in one stream never recovers — *shipped (hardware pass still owed)*
+### 28. A second device outage in one stream never recovers — *shipped (ALSA hardware pass still owed)*
 
 Found by item 27's CoreAudio hardware pass — two unplugs in one run, the first back in 190 ms,
 the second never — but it belongs to item 14 and sits in the shared `SinkRecovery`, so every
@@ -2677,7 +2677,13 @@ recoveries per stream would turn a flaky cable back into the silence this item r
 
 The regression tests (`ASecondOutageInTheSameStreamRecovers`,
 `EveryOutageAfterARecoveryStillGivesUp`, `TheDiscardedGapSurvivesTheRefillIntoTheNextOutage`)
-fail against the old helper and pass against the new one. **Still owed:** two unplug/replug
-cycles in one stream on real hardware — ALSA first, since it has the most users, then CoreAudio
-with item 27's harness.
+fail against the old helper and pass against the new one.
+
+**CoreAudio is validated on hardware.** Two unplug/replug cycles in one stream on a USB DAC,
+through item 27's harness: the sink reopened 260 ms after the first replug and 180 ms after the
+second — the second being the one that recovered never before this fix, and faster than the
+first because the refill hands it an unspent in-place reopen. Both outage gaps were reported
+rather than swallowed (drift −5554 ms mid-outage, −96 ms five seconds after the replug), and the
+run settled on the same fixed ~90 ms end-of-stream tail a clean run shows. **Still owed:** the
+same two cycles on ALSA, which has the most users.
 

docs/release-notes/0.3.0.md

@@ -0,0 +1,44 @@
+### Breaking changes
+
+- **macOS plays through a native CoreAudio backend, not PortAudio.** The macOS binary is now
+  built without PortAudio and reports `null, stdout, coreaudio`, so a device named as
+  `-o portaudio:<device>` on a Mac is `-o coreaudio:<device>` instead; run `-l` for the names
+  this host reports. The same change fixes a launch failure: the 0.2.0 macOS binary aborted at
+  startup on any Mac without Homebrew's PortAudio, and this one needs nothing installed —
+  CoreAudio, AudioToolbox and Bonjour are all part of macOS. PortAudio is untouched on Linux,
+  and a macOS build can still ask for it with `-DSENDSPIN_CLI_WITH_PORTAUDIO=ON`. (#66)
+
+### New features
+
+- **`-s` dials a typed-in address again, alongside `mdns:` discovery.** 0.2.0 removed the
+  address form; that removal is reversed by choice, because pointing a player straight at a
+  known server is worth more than matching the spec's connection model exactly. `-s` once more
+  takes a bare host (server default port 8927), `host:port`, a full `ws://`/`wss://` URL, or a
+  bracketed IPv6 literal, typed or as `server =` in a config file, and a bad address fails at
+  parse time rather than after the daemon starts. `-s mdns:[<name>]` still discovers, unchanged,
+  and `mdns:` is reserved only before the first colon — so `hifi:8927` is a host and a bare
+  `-s mdns` is a host named `mdns`. Any `-s` still suppresses the mDNS advertisement, and a
+  parse error never quotes a credential the address carried. (#68)
+
+### Fixed issues
+
+- **A second device outage in one stream recovers.** A rescan that succeeded retired the
+  recovery budget instead of restoring it, so after one unplug and replug the next outage got
+  no attempts at all: the sink discarded audio until the following track, with no log line to
+  say why. The budget is now per outage rather than per stream — a recovery refills the in-place
+  reopen and the whole rescan ladder — while a device that never comes back still gives up on
+  the same schedule as before. This sat in the shared `SinkRecovery`, so ALSA, PulseAudio,
+  PipeWire and PortAudio all had it and all get the fix. (#67)
+
+### Under the hood
+
+- **The CoreAudio sink was validated against real hardware**, not only against its tests: tone
+  accounting and DAC-offset checks at each supported format, a volume-ramp check, a
+  default-output move, and unplug/replug cycles on a USB DAC. Two findings came back onto the
+  branch before it merged — a doubled safety offset in the DAC timestamp, and a reopen that
+  waited out the backoff ladder instead of acting on the device-list change — and one, the
+  second-outage bug above, turned out to belong to every device-backed sink. (#66, #67)
+- **The macOS release leg builds with no third-party runtime dependency at all,** and CI asserts
+  the backend list it produces.
+
+**Full changelog:** https://github.com/@REPO@/compare/v0.2.0...@TAG@