[solari-ci-status]

Port of the SDK-SUITE phase 2 fixes from the private SDK source, for tag v0.2.1 after merge.

• commands.run with onStdout/onStderr no longer returns empty or truncated output when the guest's first output arrives before the cmd.start reply.
• CommandOptions::timeoutMs is a command deadline on the WebSocket path: the command is killed (SIGKILL) and the result is exitCode -1 with the output so far. New CommandHandle::waitFor(ms).

47 test cases / 350 assertions pass (built out of tree, -DCMAKE_POLICY_VERSION_MINIMUM=3.5 on CMake 4.x as before).

🤖 Generated with Claude Code

https://claude.ai/code/session_01D22yA2FD2Cm7HuYGcwAXSN

Update: also includes the change from pinetree-desktop#680: body-less POSTs (pause/resume/start) now carry an explicit Content-Length: 0. Over the HTTP/2 the real gateway negotiates these calls already worked, so for C++ this is hardening (an HTTP/1.1 connection would get 411 without it; the Rust SDK did fail in the field). Files src/http.cpp, tests/test_empty_post.cpp and its CMake entry, byte-identical to that PR. 50 test cases / 371 assertions pass.

[solari-ci-status]

ARCHI-SIGNOFF sandbox ceef3e1

Hand-port for the patch release. I compared every changed file with the merged private source on pinetree-desktop main by SHA-256: include/solari/sandbox.hpp, src/sandbox.cpp and tests/test_main.cpp (the merged private #640 and #646) are byte-identical to sdk/ there. Nothing else in the PR. The repo's CI job (the C++ build and ctest) is green on this head (1 check). Tag v0.2.1 only after the private release PR (#666) and the registries are done, and never re-point a tag.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NmYLnJPcpNnnK5pSvg7QNr

[solari-ci-status]

ARCHI-SIGNOFF sandbox 43116e0

Re-sign of ceef3e1. The delta is three files: src/http.cpp (+4 lines) and the new tests/test_empty_post.cpp are byte-identical (SHA-256) to desktop#680's head (d5fdb36b4716), and CMakeLists.txt differs from the private one as it always did but its change is exactly the one added source file (tests/test_empty_post.cpp appended to solari_tests). I built and ran that code on the private tree (50 cases / 371 assertions pass; removing the header makes 2 cases fail). Everything else in the PR is unchanged from what I signed. The repo's CI job was still running at sign-off (1 check): merge only when it is green; a red job voids this. Tag v0.2.1 only after #666's registry order, and never re-point a tag.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NmYLnJPcpNnnK5pSvg7QNr

[solari-ci-status]

ARCHI-SIGNOFF sandbox e451835

Re-sign of 43116e0. The delta is CMakeLists.txt only (+5/-1): the same pins as desktop#683, and I re-checked the values rather than trusting the port: the nlohmann/json tarball hash d6c65aca… and the doctest hash 632ed2c0… (declared twice, both carry it) are the SHA-256 values I computed from the downloaded v3.11.3 / v2.4.11 tarballs, and GIT_TAG c5a02f1066fb0fde48f80f51178429a27f689a39 is what git ls-remote gives for IXWebSocket v11.4.5. No other file changed since the sign-off I gave (src/http.cpp and the empty-POST test remain byte-identical to desktop#680). The repo's CI job was still running at sign-off (1 check, not green): merge only when it is green; a red job voids this. Tag v0.2.1 only after #666's registry order, and never re-point a tag.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NmYLnJPcpNnnK5pSvg7QNr