Skip to content

Handoff 2026-09-02 - the manuals round

Everything below is committed as 744b2b4c on xmsg-chat, except the seven untracked debug pulls in listings/ (RTLOG, HCOPY, CHATLG2) which are noise and can be deleted.

What was done, and how it was proved

Thing Where Proof
user manual DOC/manuals/CHAT-USER-MANUAL.md screens captured on D100 (FJELL) and D102 (VIDDA) with two people talking across the trunk, /who, /help, /map, /list, /tell both ways, windows, /leave
admin manual DOC/manuals/CHAT-ADMIN-MANUAL.md CHAT-MON ?, STATUS, LIST-TRUNKS, LIST-MEMBER captured on D100
build and deploy manual DOC/manuals/CHAT-BUILD-AND-DEPLOY-MANUAL.md the loop was run end to end twice today (client, then client+server)
spec updated DOC/CHAT-APP-SPECIFICATION.md /w /window /close added; "windows designed, not built" replaced
install + source floppies tools/make-floppies.ps1, dist/NDCHAT-INSTALL-2026-09-02.img, dist/NDCHAT-SOURCE-2026-09-02.img, dist/NDCHAT-2026-09-02.manifest.txt binaries pulled off D100 through the daemon (byte counts match FILE-STATISTICS, top bits intact); every file extracted back from the image and compared byte for byte
DBGUI=0 SINTRAN-CHAT/CHAT.PLNC title row reads TESTER@FJELL, no numbers
putNumber five digits SINTRAN-CHAT/CHATSV.PLNC STATUS used to print empty o45 for 6345
build stamps bumped both sources STATUS reads build S02-1340, LIST-MEMBER reads B02-1340
deployed all three machines server RT-loaded on segment 211 on D100/D102/D103, trunks all up, NDCHAT re-dumped on D102, a line said each way FJELL-VIDDA

Where the machines are now

All three run server S02-1340 on segment 211 and client B02-1340. Every trunk up. No client is joined (all exited cleanly). No sync daemon is running - the one for this session (PID 60052) was killed by hand at ~14:31 while idle; it sent no DISC, so the first daemon started next may sit on "the link has not learned the peer yet" until D100 addresses us (@COPY-FILE D19999(SYSTEM).WAKE:TXT,CHAT:CNFG on D100 - the copy fails and it still works).

History on D100's LOBBY started empty after the reload - histNext is not persisted across an RT-load (the recovery scan was reverted on 2026-09-02 because CONTINUOUS unwritten blocks are garbage). Not a regression; a known limit, now in the plan.

Not done, on purpose

  • The install floppy has not been mounted on a machine. attach fd0 <img> / mount floppy0 0 <img> live in RetroCore's console window, which I do not drive, and the DAP evaluate cannot run console commands (connecting also takes the CPU and leaves it paused on disconnect). Ronny chose to skip it for now. Admin manual 3.1 says UNVERIFIED; the SINTRAN side (ENTER-DIRECTORY, COPY-FILE) is written from convention, not measured.
  • CHATRT:MODE on the floppy has never been run as a MODE file; its step list is rt-load.ps1's, which is proved.

Found while photographing - all in DOC/CHAT-PLAN.md, items 1-9

  1. After /nick the client's corner and the "sent" half of a private window keep the login name. Other end sees the right name.
  2. Header shows LOBBY@trunk for a plain LOBBY.
  3. A topic set on FJELL did not reach VIDDA.
  4. /leave on a screen terminal shows nothing.
  5. /list counts local seats only.
  6. putNumber - fixed and deployed.
  7. Stamps hand-set, never bumped - bumped; a check that cannot be forgotten is still wanted.
  8. planc-lint.py treats all files on its command line as one link set (CHAT + CHATSV draw a false inBuf collision). Lint per link set.
  9. nd-deploy.ps1 stages into sync-out, the daemon watches sync-relay; its fallback stops every RetroCore. Do not use it.

Traps met today, for the next person

  • FILE-STATISTICS CHAT:PLNC lists every CHAT* file and ;1 does not narrow it - the one you want scrolls off; read the scrollback.
  • Asking the daemon to pull a file a MODE file has just deleted is refused with SINTRAN error 110 and reads like a transport fault.
  • ndtool -x -p strips the parity bit from binaries too - extract text and binaries separately.
  • Feeding the linter both programs at once is not a link set.