VIPIRC 0.3.1 - EVIDENCE INDEX

The release report is ../VALIDATION.txt. It distinguishes optimized client
measurements, instrumented runs, standalone engine fixtures, and interrupted
or superseded attempts. Raw logs must be read with their matching JSON and
this index. A version string alone does not identify a build; use SHA256.

FINAL OPTIMIZED BUILD

SHA256 44af2f425eee4f8e5ce4dbee71f4232123a5f89cbc89285164a06c7f6fbaf50c
final-build-031.json: executable, manual, compiler and runtime-source hashes.
strict-final-031.log: empty output means the strict syntax command was clean.
build-final-031.log: final optimized compilation/link commands.
final-gates-031.json: recorded execution-tool exit statuses and environments.
check-final-031.log: full normal regression target.
sanitize-final-031.log: full ASan/UBSan target, leak checking enabled.

The sanitizer compiler emits one reviewed -O1 format warning about the
inline Buffer.name array. This is distinguished from runtime sanitizer
reports; the optimized all-source -Werror check is clean.

FINAL ADDITIONAL WORKLOADS

ui-stress-final-031.json/.log: full optimized seeded PTY stress.
ui-stress-sanitize-final-031.json/.log: same suite with ASan/UBSan and leaks.
ui-fix-review.txt: root causes, focused reproductions, and scope limits.
ui-stress-source.patch: reviewed UI change against the 0.3.0 baseline.
ui-wide-caret-fixed.json/.log: focused wide-glyph checks on a private build.
fd-pressure-final-031.json/.log: exact-release FD limits 5, 8, and 16.
fd-pressure-poll-root-cause.log: isolated old poll-count failure.
soak-final-1200.json/.log: intended 1,200-second verified-TLS soak, interrupted
by the user after the 720.2-second progress checkpoint. The JSON is empty;
this run is incomplete and NOT counted as passed. See
soak-user-interruption-031.json for the recorded observations and scope.
No new final idle/throughput benchmark was run after the interruption.
benchmark-030-previous-release.json in the archive is historical 0.3.0 data,
not a measurement of the current executable.

app-faults.log and app-faults-sanitize.log: exact App allocation balance,
rollback and 12,000 injected ENOMEM failures in each run.
app-faults-build.log: normal harness build.
app-faults-sanitize-reproduce.sh: private build and instrumented reproduction.

dcc-protocol-stress-evidence.json is authoritative for DCC figures. Its
production-source hashes bind the tested engine/parser to the release.
The normal 90-second live run and final leak-enabled 60-second live run
have complete summaries and successful exit evidence. Earlier rough
lifecycle labels were replaced with exact counts without changing workload;
final strict-build/smoke checks are also retained. DCC RSS is harness RSS,
not application RSS. Timer injection is distinguished from elapsed waiting.
dcc-protocol-stress-reproduce.sh builds in a private mktemp directory.

transport-config-storage-normal.log records the final normal storage fixture,
including the real short write and ancestor-directory race.
transport-config-lifecycle-leak-final.json/.log records the final network
and storage repeat with explicit leak/undefined-behavior options, production
hashes, and child exit status. This resolves an evidence gap in the earlier
instrumented logs: their runtime sanitizer options were not recorded.
transport-config-storage-sanitize.log and -lifecycle-sanitize.log retain
those earlier instrumented passes, without an independently verified claim
of leak detection. See transport-config-sanitizer-environment-status.json.
transport-config-network-normal-summary.json is explicitly reconstructed
from captured tool output because its original stdout was not redirected.
transport-config-client-*.log and -config-*.log record their component tests.

STATIC ANALYSIS, MANUAL, AND RELEASE BUILD

gcc-analyzer-triage.json: partial static-analysis scope and diagnostic review.
gcc-analyzer-*.log: compiler output; complexity cutoffs are retained.
gcc-analyzer-dup2-reproducer.c/.log: minimal compiler-model false positive.
license-original-020-comparison.json: exact supplied-license comparison.
man-031-preflight.*: groff source hash, command, clean diagnostics and render.
release-clean-031.json/.log: isolated byte-identical rebuild, installed file
modes/content, CLI/manual checks and staged uninstall. Runtime-source hashes
are compared again with the final source archive during packaging.

SUPERSEDED, PRE-FIX, OR INCOMPLETE ATTEMPTS

baseline-*.log: released 0.3.0 baseline, not the final 0.3.1 executable.
build-031.log, check-031.log and sanitize-031.log: an earlier 0.3.1 build
before the final wide-caret and low-descriptor fixes. Use *final-031 instead.
ui-stress-031-before-wide.json/.log: earlier optimized suite before adding
the per-step visible-caret assertion; it did not establish the later fix.
ui-wide-caret-repro-before-fix.*: the real redraw failure that led to the fix.
ui-mouse-floods-031.*: focused earlier mouse-flood checks; final suite repeats
them with both final builds.
fd-pressure-before-fix-031.log and fd-pressure-031.*: original harness/error
investigation; not final passing evidence.
soak-smoke.*: short smoke, not a 20-minute longevity result.
soak-superseded-1200.* and soak-superseded-interruption.json: intentionally
interrupted earlier build, not a completed 1,200-second pass. KeyboardInterrupt
records the controlled interruption. Its JSON may be empty because no final
summary was emitted before cleanup.
dcc-protocol-stress-asan-lsan-incomplete-attempt.log lacks a complete final
summary and is classified as inconclusive even though a tool reported exit 0.
Earlier DCC files with no-lsan in the name intentionally disabled leak
checking and are not used to claim a leak-enabled pass.

All passing claims require a completed matching run and successful exit.
No finite test suite or partial static analysis proves absence of all bugs.
