RunCPM: upgrade to 6.9 and consolidate transports onto one shared core

[peterkaczorowski]

Summary

Updates the vendored RunCPM engine to 6.9 and refactors all CP/M transports onto one shared stateless core.

The 6.9 bump and the core refactor are in the same PR by necessity: 6.9 has substantially more static global state than 6.5, and each CP/M transport TU (siocpm, drivewire, iwm, rc2014, rs232) carries its own private copy via internal linkage. Five copies of 6.9's larger state overflow the ESP32 DRAM .dram0.bss segment by ~58 KB and the firmware will not link. Consolidating to a single engine instance is the prerequisite that makes 6.9 fit.

What changed

• runcpm_core.cpp is now the only TU that compiles the Z80/CP/M chain (external linkage). runcpm_session.h exposes runcpm_console_ops; abstraction_fujinet_core.h holds the shared disk/SD section.
• The transports (siocpm, drivewire/cpm, iwm/cpm, rc2014cpm, rs232cpm, network-protocol/CPM) become thin console back-ends that plug into the shared core through runcpm_console_ops, guarded by a single-instance interlock.
• A read-handle cache collapses RunCPM's per-128-byte-record open/seek/read/close on the SD card into one cached read-only handle shared by _sys_readseq/_sys_readrand.
• Symbol-collision renames so the core TU can pull in bus headers on Apple/COCO targets: RunCPM Command struct → CcpCommand (collides with FujiNet class Command); CPM macro → handled in globals.h banner.

Test plan

• ESP32 fujinet-atari-v1: ./build.sh -b links clean — RAM 36.8%, no dram0_0_seg overflow (vs. overflow-by-58 KB without the consolidation).
• SIO 'G' CP/M and N:CPM:// regression on hardware.

---

This is PR 1 of 3 of a dependency stack that replaces #1371. Merge first; PRs 2 and 3 build on it.

🤖 Generated with Claude Code

[peterkaczorowski]

Thanks for the review @FozzTexx — both points addressed.

"Touching a lot more than just upgrading runcpm"

It unfortunately can't be a pure version bump. 6.9's larger static state, replicated across all five transport TUs (SIO, IWM, DriveWire, RC2014, RS232), overflows the ESP32 .dram0.bss segment by ~58 KB — it won't link. The fix is to stop compiling RunCPM five times: the transports now share a single stateless core (runcpm_core.cpp + runcpm_session.h) and each bus just provides a small console back-end. So the 6.9 upgrade and the shared-core consolidation are one inseparable change — you can't land 6.9 without it.

To make that reviewable I split the PR into two commits:

1. Update vendored RunCPM to 6.9 (verbatim upstream) — a clean drop of the 10 stock upstream files, byte-for-byte identical to MockbaTheBorg/RunCPM @ 6.9. You can review this one in isolation by diffing against upstream; it should be empty.
2. FujiNet: integrate RunCPM 6.9 onto a single shared core — all the FujiNet-specific work: the shared core, the linkage shim, the name-collision renames, and the transport back-ends. This is the commit to actually review.

"Walls of comments"

Stripped them throughout — the integration commit is now net -1137 lines. I also deleted three abstraction headers that the shared-core refactor made dead code.

Build stays green at 36.8% RAM (fujinet-atari-v1). The two stacked PRs (#1392 telnet console, #1393 bar commands) are rebased on top of this.

[peterkaczorowski]

Superseded by a cleaner split: #1401 (consolidate bus CP/M devices onto a shared cpmDevice base, on the current 5.8 engine) followed by #1402 (engine bump 5.8 → 6.9 on that base). Closing in favour of those.