[aceRage]

Description

Phase 1 of the Creality send-gap stack (research C1+C2+C5). Closes the “Upload and print reports success but never starts” class of bugs on K2-platform printers, and surfaces start failures instead of swallowing them.

Does not affect Snapmaker printers. Needs a Creality K1/K2 hardware smoke before merge (not a U1 check).

Review 1/3 (F1, F2)

• F2: tcp_stream::expires_after does nothing on synchronous Beast. CrealityWsClient now runs every resolve / connect / handshake / write / read as an async op with io_context::run_for, and closes the socket on expiry. Shared by start_print, the Phase 2 CFS probe, and Phase 3 sync.
• F1: when /info is empty or unknown, resolve_model() falls back to the preset printer_model (Creality K1C → K1C) so a K1 never regresses to /mnt/UDISK.

Re-review 1/3 (B3, F15)

• B3: ready_to_start() refuses to start an op when budget() <= 0 or the client is cancelled, and run(timeout_ms <= 0) still cancels then ioc.restart(); ioc.run() so every path drains before the locals go out of scope.
• F15: resolver.cancel() cannot interrupt getaddrinfo; the DNS wait is bounded by the OS. Literal IPv4/IPv6 skip resolve via asio::ip::make_address.

Re-review 2/3 (B5)

• B5: cancel() is now UI-thread safe. It only stores cancelled and net::post()s cancel_socket() onto this client’s io_context. It does not call ioc.stop() and does not touch open / ws off-thread. open() no longer clears cancelled, so a cancel before open() is not lost. open is std::atomic<bool> (worker-only writes). Every run() path still drains (B3): timeout, already-cancelled, and cancel-during-run_for all on_timeout() then ioc.restart(); ioc.run().
• Catch2: cancel from another thread before open(), and during a never-accepting connect.

Upstream sources

• OrcaSlicer Creality K1-family CFS support (K1 SE) — follow-up to #13752 OrcaSlicer/OrcaSlicer#14089 (CrealityPrint.cpp is_k1_family(), /info model query, omit multipart path on CFS models, start_print bool + error_fn).
• OrcaSlicer Http::get_host_from_url — this fork’s equivalent is Http::url_host (scheme and port stripped).
• Profile host_type sync from upstream Updated profiles for all Creality K-series printers (+CFS-c support) and Creality SparkX i7 support OrcaSlicer/OrcaSlicer#13947/Creality K1-family CFS support (K1 SE) — follow-up to #13752 OrcaSlicer/OrcaSlicer#14089: crealityprint + printer_agent: crealityprint on K1 / K1 SE / K1C / K1_CFS-C / K2 Plus.

What was adapted

• Network-free helpers live in CrealityPrintModel.hpp so Catch2 can pin them without a printer. Not a cherry-pick of upstream CrealityPrint.cpp.
• Kept this fork’s PrintHost::apply_tls on /info and upload (upstream still uses #ifdef WIN32 ssl_revoke_best_effort only).
• start_print still sends only opGcodeFile (CFS colorMatch / multiColorPrint is Phase 2). Root is /usr/data for K1-family (including preset names) and /mnt/UDISK for everything else (K2 / K2 Pro / K2 Plus / SPARKX i7).
• WebSocket connect uses TEXT frames and Http::url_host so a print_host written as http://host:80 resolves.
• A failed start now calls the job’s error_fn and returns false from upload().
• Profiles: only host_type / printer_agent added (or octoprint → crealityprint on K1_CFS-C). No other preset fields touched. Cross-checked against Profiles: Creality preset sanity fixes #319 (Creality sanity: default process / max_layer_height) and Profiles: refresh vendors to upstream Orca + preset sanity fixes #320 (Creality was out of scope). Vendor version 02.03.02.05 → 02.03.02.06.

Expected conflicts

• QIDI Box: send the filament-to-slot mapping before starting a print #610 / QIDI send: offer QIDIStudio's print options (Box, leveling, timelapse, polar cooler) #611 / Happy Hare and AFC slot reading for the Moonraker send dialog #612: none on this PR (they do not touch CrealityPrint.cpp or these profiles).
• Native Moonraker print host (upload without octoprint_compat) #613: none.
• ElegooLink: restore the Elegoo send dialog so Upload and print sends a valid start command #594: already merged; no overlap.
• Later phases in this stack merge this branch upward.

Untested

• No Creality hardware in this environment.
• slic3rutils_tests was not compiled here. Catch2 cases were added to the existing tests/slic3rutils/printhost_devices_tests.cpp (no CMake test edits).
• No G-code output changes.

Tests

Catch2 v2 in printhost_devices_tests.cpp:

• K1-family → /usr/data/..., omit upload path
• F008/F012/F021/F022 → /mnt/UDISK/...
• empty / unknown /info falls back to preset printer_model
• ws_host strips scheme and port
• parse_model_from_info is hostile-JSON safe
• expired total budget fails before any I/O (B3)
• local acceptor that never accepts times out and drains (B3)
• cancel from another thread before open() and during a never-accepting connect (B5)

Smoke checklist (Creality hardware)

1. K1C (or K1 / K1 SE): Host Type CrealityPrint, Upload — file lands under /usr/data/printer_data/gcodes.
2. K1C Upload and print — print starts; a refused start shows an error, not success. An unreachable printer must not hang (F2 deadline).
3. K2 / K2 Plus / K2 Pro / SPARKX i7: Upload and print — print starts from /mnt/UDISK/printer_data/gcodes (the old /usr/data path is the C1 bug).
4. print_host as http://<ip> and http://<ip>:80 still connects (scheme/port stripped).
5. Plain Upload (no print) still succeeds.
6. A K1 whose /info omits model still uses /usr/data (preset fallback).
7. Calibration toggle / CFS mapping are not in this PR (Phase 2).