master does not configure, compile, link or pass tests after the 09-07 batch merge

[Kartikey1306]

Symptom

master has not configured, compiled, linked or passed a test since the 09-07 batch merge. Every CI — eBoot, eBoot Build & Test, CodeQL and nightly run on master is red, and eos's simulation and nightly "Upstream drift" jobs — which build eBoot at master — are red with it.

Verified on b429354: configure aborts on a duplicate eboot_test_fdt_loader target; with that removed, core/sha512.c fails ×5 on eos_sha512_ctx_t having no member count and core/ed25519_verify.c:303 on a redefined scalarbase; with those fixed, eos_secure_boot, eos_secure_boot_lock_debug and eos_boot_log_append are undefined at link.

Cause

Twenty PRs were merged in ninety minutes on 09-07, each replayed onto a master its author never saw. #55 (written 08-28) restored a stale sha512.c and dropped three sources from eboot_core; #94 and #105 both repaired the Ed25519 verifier and both added scalarbase(); #84/#85 registered the same test target twice; #104 and #103 disagree on the install path's verification order.

Fix

#115.

[Kartikey1306]

Status as of 2026-09-17, so this issue reflects the current state rather than the one it was opened on.

• Repair: fix: repair master after the 09-07 batch merge #115, head e152d8e — MERGEABLE, 26 checks passing, 0 failing, blocked only on a review. The head has not moved since the last re-approval request on the PR.
• Master: still red since the 2026-09-07 batch merge: host configure/build (sha512.c no member count, scalarbase redefined) and EoSim Sanity on the windows-latest legs (09-16, 09-17). Every run of those workflows on master since then has failed; nothing else has touched the cause.
• Blocked behind it: fix(keystore): the compiled-in dev trust anchor was not a point on the curve #116, fix(sha512): drop a carry term that is always zero #123, ci: grade only open pull requests against the linked-issue policy #125, fix(recovery): refuse an unprovisioned secret, and a write into an unmapped slot #127, fix(stage0): refuse a stage-1 image the flash could not read #129, ci(book): Eisvogel 3.5.1 for pandoc 3.6.1, and let the workflow test itself #135 — each stacked on fix: repair master after the 09-07 batch merge #115's head and each green on its own diff.

Merging #115 first clears master and lets the rest rebase to their own single commits. Nothing further is needed from the author side; the PR body and comments carry the verification.