diff options
Diffstat (limited to 'ocaml/FINDINGS.md')
| -rw-r--r-- | ocaml/FINDINGS.md | 113 |
1 files changed, 113 insertions, 0 deletions
diff --git a/ocaml/FINDINGS.md b/ocaml/FINDINGS.md new file mode 100644 index 0000000..397924c --- /dev/null +++ b/ocaml/FINDINGS.md @@ -0,0 +1,113 @@ +# Critical Boot System Findings (2025-01-06) + +## Summary +We discovered that the ivory.pill file structure is `["ivory" ARVO_CORE]` where the tail is a CELL containing the complete Arvo core, NOT atom 0. However, the lifecycle formula `[2 [0 3] [0 2]]` FAILS when executed on this Arvo core. + +## Detailed Findings + +### 1. Ivory Pill Structure +**Verified by:** `test/examine_ivory.ml` + +``` +Tag: ivory +Tail type: CELL + Head type: cell + Tail type: cell +✓ Has slot 2 and 3 +``` + +The ivory.pill file contains: +- Tag: The atom "ivory" +- Tail: **A complete Arvo core (CELL)** - NOT null/atom 0! + +### 2. Lifecycle Formula Execution Test +**Tested by:** `test/test_two_stage_boot.ml` + +Running `[2 [0 3] [0 2]]` on the Arvo core: +- ✓ Step 1: `*[core [0 3]]` succeeds → returns cell +- ✓ Step 2: `*[core [0 2]]` succeeds → returns cell +- ✗ Step 3: `*[slot3 slot2]` **FAILS with Nock Exit** + +### 3. What This Means + +The ivory.pill file's Arvo core **cannot be bootstrapped** using the lifecycle formula in our implementation! This leads to two possibilities: + +**Possibility A:** The embedded ivory in C Vere is DIFFERENT from ivory.pill +- Embedded might be `["ivory" 0]` or some simpler structure +- ivory.pill file might be for a different use case + +**Possibility B:** C Vere's `-B solid.pill` path doesn't use ivory at all +- Your log showed `u3v_life: eve=null` during solid pill boot +- Maybe solid pill boot skips the ivory pill entirely? +- The `u3v_boot_lite()` code in mars.c was COMMENTED OUT! + +## Questions to Answer + +1. **Does `-B solid.pill` load ivory.pill at all?** + - Need to trace mars.c boot path for `-B` flag + - Check if embedded ivory is even used + +2. **What is eve=null in your C Vere log?** + - If eve is null (atom 0), how does `[2 [0 3] [0 2]]` work on it? + - Our test proves it CANNOT work on atom 0! + +3. **Is ivory.pill the right file to use?** + - Maybe there's a different pill for bootstrapping? + - Or maybe solid.pill contains everything needed? + +## C Vere Test Results + +**Test Created**: `vere/pkg/noun/life_test.c` +**Built with**: `zig build life-test` + +**Result**: +``` +Testing lifecycle formula [2 [0 3] [0 2]] on null + +Formula: [2 [0 3] [0 2]] +Subject: 0 (null) + +✗ FAILED! +Error: : %exit + +This confirms the formula CANNOT work on atom 0! +``` + +**PROVEN**: The lifecycle formula `[2 [0 3] [0 2]]` **FAILS** on atom 0 in C Vere! + +## The Contradiction + +Your boot log showed: +``` +u3v_boot: processing 10 events +u3v_life: eve=null (atom=yes cell=no) +u3v_life: completed successfully +``` + +But our C test PROVES this is impossible! The lifecycle formula cannot succeed on null! + +## Possible Explanations + +1. **"eve=null" doesn't mean the lifecycle formula is called on null** + - Maybe it's just logging what the original `eve` was before processing? + - The actual u3v_life call might receive something else? + +2. **The log is from a DIFFERENT u3v_life call** + - Maybe there are TWO separate calls to u3v_life? + - One with ivory pill's core (succeeds) + - One that's never actually executed (logged as null)? + +3. **There's special case handling we're missing** + - Maybe C Vere checks for null and skips u3v_life entirely? + - The log might be misleading? + +## Next Steps + +1. Add MORE detailed logging to C Vere's u3v_life to see: + - Exact eve structure being passed + - Whether formula actually executes + - What the result is + +2. Trace where u3v_life is called from during `-B solid.pill` boot + +3. Check if there's special handling for null in u3v_boot or mars.c |