c02fed140ffcb0f18db6bc3c24bf978f13b9d6ab
4
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Robert Allan JamesandClaude Sonnet 5
|
c02fed140f |
Fix ISA-tagging: two real bugs found live, campaign quit for a fresh run
FABRIC-3.md §XXXV.13. Quit the running amd64 campaign directly on request once §XXXV.12's build unblocked live testing -- first live check found the ISA-tagging mechanism completely non-functional, in two independent ways, both only caught by checking an actual printed value rather than "it resolves without error." Bug 1: per-isa-doe.4th's strategy of redefining MU-EMIT-TRIAL-MARKER never worked, structurally, from the moment it was written. This FORTH compiles a colon definition's word calls to a direct, early-bound reference at compile time -- MU-RUN-TRIAL's call to MU-EMIT-TRIAL-MARKER bound when multiuser-doe.4th loaded, before per-isa-doe.4th's later redefinition of the same name ever existed. Every trial the running campaign persisted had an empty isa tag, not "amd64". Fixed by moving ISA-tag storage (MU-ISA-TAG-BUF/LEN/!) and the one true MU-EMIT-TRIAL-MARKER into multiuser-doe.4th itself, reading a shared variable at run time instead. per-isa-doe.4th shrank from 3 blocks to 2 (Block 5118 freed) -- fixing this recovered namespace, didn't cost any. Bug 2: found immediately while verifying the fix. MU-ISA-TAG!'s CMOVE call was copied from the original PER-ISA-TAG! it replaced, which had its own latent stack-order bug -- passed the ISA string's length as the copy source address instead of the real address, silently copying blank/garbage bytes with no error. Never caught before because §XXXV.8's own verification only checked that the entry words resolved, not that the tag they set was actually correct. Fixed the stack order; the old buggy word no longer exists anywhere (removed with Bug 1's fix, not patched in place). Verified live, together, one boot: MU-ISA-TAG-BUF readback -> "amd64" (was blank), full MU-EMIT-TRIAL-MARKER with all 6 fields set manually -> exact match on every field, VM-ERROR? -> 0, DOE-RESUME-COUNT correctly isolates by ISA (amd64->1, riscv64->0 in the same ring). Clean BYE, zero leaked processes, synthetic test write reverted before committing. Preserved logs/CSVs from the killed campaign and both verification boots per this repo's own convention -- never delete these. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BWpNjdwPtFLuVLaAq44L9K |
||
|
Robert Allan JamesandClaude Sonnet 5
|
51940f49d0 |
Add per-slot CRC + campaign crash-resume (DOE-RESUME-COUNT)
FABRIC-3.md §XXXV.12: closes two real gaps found by walking the disk-to- PDF pipeline end to end. (1) doe_trial_slot_t had no integrity check of its own -- 64 slots share one devblock via a read-modify-write, so a crash mid-write to any slot could silently corrupt its siblings with nothing to catch it. Added an 8-byte CRC (same discipline the control header already had), computed/verified the same way. (2) The campaign itself had no crash-resume -- a reboot always restarted the whole 18-trial loop from index 0 even with good persisted data already on disk. Since the trial matrix is fully deterministic from (seed, n-reps), resuming needs no persisted "which trials ran" state, just a count: DOE-RESUME-COUNT (new kernel primitive, same pattern as DOE-PERSIST-TRIAL) walks the ring counting CRC-valid records for a given ISA; MU-EXEC-CAMPAIGN's signature changes to accept a start-idx and seeds MU-RUN-ID/the DO loop from it. Each RUN-*-CAMPAIGN entry word now calls DOE-RESUME-COUNT before launching. Both block-namespace-checked to fit existing headroom before committing to the design -- no new blocks needed. Real bug caught by the build, not shipped: an implicit compiler-inserted alignment byte before slot_crc silently broke the hand-computed trailing pad size (doe_trial_slot_size_check failed to compile). Fixed with an explicit alignment field, matching log_slot_t's own zero-implicit- padding discipline. extract_doe_region.py now verifies both CRCs with the REAL CRC-64/XZ algorithm from block_subsystem.c's compute_crc64() (ported and verified bit-for-bit against a native C reference this session), replacing §XXXV.11's incorrect guessed implementation. Verified: clean compile on all 3 ISAs (amd64/aarch64/riscv64 -- amd64 verified via isolated build only, not boot, since the real per-ISA campaign is currently live on the only QEMU instance this project's own rule allows), both capsules pass mkcapsule --lint, CRC64 bit-exact via native reference. Live boot-verification explicitly deferred until that campaign finishes or is interrupted -- not claimed as tested when it wasn't. Found (not missed) a real compatibility consequence: the running campaign's data is in the old pre-CRC format, so it reads as "corrupt" under the new extractor and DOE-RESUME-COUNT will resume it from 0, not mid-campaign -- no data loss, cheap redundant re-run of ~1-2 trials. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BWpNjdwPtFLuVLaAq44L9K |
||
|
Robert Allan JamesandClaude Sonnet 5
|
2da746c01c |
Build DOE-PERSIST-TRIAL: block-backed DoE campaign persistence, live-verified
FABRIC-3.md §XXXV.3/.4 designed this and it was never built -- two live campaigns got launched relying entirely on a live serial-log tail before this was caught. doe_region.c/.h mirrors log_region.c's own growable-ring pattern (own devblock_from_top fence at 98, past log_region's ceiling at 97) to persist one trial-summary record per completed trial straight to disk/artemis.img, surviving past QEMU exit with no host required. Verified end to end: booted amd64, called DOE-PERSIST-TRIAL at the console, clean BYE, then read the raw disk bytes back with QEMU fully dead -- confirmed the exact values passed in. Wired into both multiuser-doe.4th and per-isa-doe.4th's trial markers. Clean build, zero new warnings, all 3 ISAs; mkcapsule --lint passes both edited capsules. FABRIC-3.md §XXXV.10 documents the fix and the root-cause correction: two campaigns were launched on an unbuilt persistence design before this was caught and fixed. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BWpNjdwPtFLuVLaAq44L9K |
||
|
Robert Allan JamesandClaude Sonnet 5
|
8b8e8e1405 |
Build capsules/per-isa-doe.4th: per-ISA campaign driver (N=3, 18 trials/ISA)
Thin wrapper over multiuser-doe.4th's own MU-EXEC-CAMPAIGN, which already implements the ratified 6-cell factorial generically -- adds only what FABRIC-3.md §XXXIV.4's per-ISA pivot needed: one fixed seed per ISA, N=3 reps/cell=18 trials/ISA (ratified over §XXXV.5's own N=2 safety-margin recommendation, Captain Bob's explicit choice), an ISA tag threaded into the trial marker so records from all 3 ISAs' sequential campaigns can be told apart on the shared disk/artemis.img, and three single-word no-argument entry points (RUN-AMD64-CAMPAIGN/RUN-AARCH64-CAMPAIGN/ RUN-RISCV64-CAMPAIGN). Blocks 5117-5119 -- the only 3 remaining in the [2048,5120) user block namespace; mkcapsule --lint caught the first draft going one block past the ceiling. Verified live, load-only: both capsules load with zero UNKNOWN WORD/redefinition/error lines, all 3 entry words resolve, constants compile correctly. No campaign trial has been run yet -- build+verify only, per explicit scope for this pass. FABRIC-3.md §XXXV.8/.9 records the decision and verification. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> |