ABI choice 5: callers reach a replacement through a jump written at the old entry

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
This commit is contained in:
Claude
2026-11-19 14:58:39 +00:00
parent 5d9737223b
commit 07f86d94c8
+5 -5
View File
@@ -171,11 +171,11 @@ capsule-blocks.md` §5), and Artemis keeping what is kept (ruling 3).
the train it imports from: (its capsule hash, the word's name). If that train is rebuilt or updated,
its hash changes, so its importers are rebuilt too, and a live update replaces the whole dependent set
of trains together. Placement refuses an import whose hash is not present.
5. **How callers reach a replacement:**
- (a) the old entry's code is overwritten by a `jump` to the new: one cell, every caller follows, and the
old code is unreachable but for that cell;
- (b) the kernel's call hook redirects calls to the old address, from a table;
- (c) callers are re-patched from records kept at placement.
5. ~~How callers reach a replacement~~ **Ruled 2026-11-19: "Jump at the old entry".** At activation, each
replaced word's first code cell, at its xt, is overwritten with a `jump` to the new word: one cell per
word, and every caller, compiled at run time or not, follows at once. The old code stays in place,
unreachable but for that cell, until it is retired. The new word must be within the jump's reach from
the old entry (27 bits in slot 0, which covers a node's memory), or activation is refused.
6. **The safe boundary:** how the system knows no node is inside the old word. It could scan return stacks
at a heartbeat for return addresses in its range, or wait until the node is idle at its prompt.
7. **Proofs:** which of ADR §5's properties come first in Isabelle/HOL. Placement's fail-closed checks and