Atlas 286 status
Here's where things stand. The labels mean the same as everywhere else: [I] (measured on real hardware) real hardware, [E] (measured in the 86Box emulator) the 86Box emulator, [U] (measured on the Unicorn CPU model) a CPU model.
Milestones, newest first
| Milestone | Where it's at |
|---|---|
| M18 | In progress. I'm working through the fixes from the 1.0 gap list one at a time, each with its own off switch and its own failing test. None of it has been on real hardware yet; that happens in one big bench session at the end. |
| M17 | Passed on the One ROMs on 25 October 2026 [I] (measured on real hardware): ten cold boots, ten warm boots, a clean ATLTEST run and five boots with HIMEM, all good. That same morning E1 showed up once in twenty HIMEM boots, plus once more right after. The release note is final and waiting for me to tag it. Windows 3.1, the floppy check on A: and a real Ctrl+Alt+Del move on to M18. |
| M16 | A big tidy-up of the source, proven to build exactly the same bytes as M15 apart from the build stamp. No tag of its own. |
| M15 | Burned to an EPROM pair on 18 October 2026 [I] (measured on real hardware), wrapped up and tagged the next day. |
| First boot | The first DOS boot on the from-scratch Atlas, on the PB686, 5 October 2026 [I] (measured on real hardware). |
What's left before a 1.0 release candidate
9 gates, 9 open. Read from the Atlas 1.0rc gate list as it stood on 2026-10-25.
| # | Gate | State |
|---|---|---|
| 1 | The provenance rulings are signed, and an audit of the release tree comes back clean. | OPEN |
| 2 | The release is burned to a real EPROM pair and passes the confirmation tests, matched against the same tests on the One ROM emulators. | OPEN |
| 3 | Drive IDENTIFY data is on record for three more CF cards, one of them the IBM 5170's, before the M18 burn. | OPEN |
| 4 | E1 and F11 stay listed as open. Nothing is called a fix for either without evidence from real hardware. | OPEN |
| 5 | CD-ROM boot runs on a real ATAPI drive, or the 1.0 docs say plainly that it has only been tested in the emulator. | OPEN |
| 6 | The IBM 5170 passes its bench test. Until then the docs claim only "IBM 5170 with an IDE/CF controller". | OPEN |
| 7 | The automated smoke-test baseline is frozen at the end of M18, with any out-of-date rows fixed or written down. | OPEN |
| 8 | An independent check compares the release images with vendor BIOSes and passes, with matches allowed only in the standard IBM interface tables. | OPEN |
| 9 | ATLASTST is built: a small DOS program a tester runs on their own board to produce a compatibility report. | OPEN |
The EPROM burn and the provenance rulings are mine to do. Until every one of these is closed there's no release candidate and nothing to download. Sadly I can't just point at the list and say "make it so".
