Record bidirectional native lifecycle acceptance and review
This commit is contained in:
@@ -0,0 +1,55 @@
|
||||
# Closed lifecycle checker and footer gates
|
||||
|
||||
Run `python verify.py` with Python 3.10 or newer. Verification uses only these
|
||||
package files and the standard library; it does not execute archived source,
|
||||
binaries, Cargo, Git or network requests. `prepare.py [CHECKOUT]` reproducibly
|
||||
packages the explicitly named raw captures if those local originals are present.
|
||||
The input capsules are source snapshots, not a hermetic toolchain distribution.
|
||||
|
||||
Ten closed Cargo runs are preserved: footer-format-1, footer-windows-build-1,
|
||||
footer-windows-build-2, footer-clippy-1, lifecycle-producer-build-1,
|
||||
lifecycle-producer-generate-1, lifecycle-producer-build-2, footer-windows-build-3,
|
||||
footer-format-2 and lifecycle-producer-format-1. Each remains bound
|
||||
to its original command, source inventory, helper, logs and receipts. The first
|
||||
footer build has no JSON executable artifact records; the second captures them.
|
||||
Recorded HEAD d0b7c28 does not identify uncommitted footer/producer source bytes;
|
||||
the archived per-run input manifests do.
|
||||
|
||||
The three final gates use the same 715-file manifest, SHA-256
|
||||
`b5a62817a1b5df936b9d759241bdfc4e8283901c1d4b540b9e494995e1544d8d`.
|
||||
Compared with producer build2's captured inputs, only the footer literal changes
|
||||
from "changes" to "reviewed items". The verifier checks that exact byte delta.
|
||||
The final GUI build and both format checks pass. Earlier lint and oracle captures
|
||||
retain their original input scope; they are not presented as reruns of the final
|
||||
wording. The checker source is unchanged by that label refinement.
|
||||
|
||||
Producer build1 and its generation succeeded, but that older checker incorrectly
|
||||
rejected expected engineering warnings during saved-file admission. Build2
|
||||
contains the corrected admission logic. The 40 PASS falsifier trials are bound
|
||||
to build2's exact checker artifact receipt and producer source, not to build1.
|
||||
Build success alone does not demonstrate checker correctness. The earlier native
|
||||
admission failure is outside these selected captures, not fabricated as a Cargo
|
||||
failure or silently converted into successful native evidence.
|
||||
|
||||
The oracle archive retains all original fixtures, repeated generation outputs,
|
||||
mutated specimens, identity receipts, copied build evidence, runner/source,
|
||||
stdout/stderr and per-trial results. The 40 expected outcomes include accepted
|
||||
original/deleted/recreated/relinked states and deliberate rejection cases for
|
||||
retained prose/geometry corruption, source/trace identity drift, missing persisted
|
||||
IDs, absent identity receipts and receipt overwrite. These are synthetic checker
|
||||
tests, not native interaction. A PASS summary includes expected negative exits.
|
||||
|
||||
`manifest.json` maps every logical raw path to a stored member. Only byte-identical
|
||||
inputs.json/inputs.zip files are deduplicated; exact run mapping is retained.
|
||||
`files.json` pins outer files including the reproducible packager. Outer text is
|
||||
UTF-8 LF; all raw archived bytes are preserved. The verifier checks exact safe ZIP
|
||||
membership, hashes, nested input contents/counts, helpers, timestamps, receipt and
|
||||
binary-pin relationships, repeated fixture equality, watched specimen identities,
|
||||
all 40 outcomes and their recorded rejection text. It rereads evidence rather
|
||||
than rerunning the model checker, and does not authenticate historical execution.
|
||||
|
||||
Native trial1 lacks a closing exit receipt and is excluded. No successful native
|
||||
deletion, Undo, relinking, save/reopen, or full lifecycle acceptance is claimed.
|
||||
No runtime binaries, credentials or whole user profile are included. Recorded
|
||||
absolute paths are historical labels; verification maps them to archive members
|
||||
without accessing the original machine. Binary hashes remain receipt observations.
|
||||
Binary file not shown.
@@ -0,0 +1,22 @@
|
||||
{
|
||||
"evidence.zip": {
|
||||
"bytes": 14254708,
|
||||
"sha256": "663bf83db4ead276063b321fd5e9857e29d2305e6046c2e87dbbac87a366b19c"
|
||||
},
|
||||
"manifest.json": {
|
||||
"bytes": 79071,
|
||||
"sha256": "5d2be3d68206871a560dd1a83064ff779e8aab6e11302d09aaf1a655502dd38e"
|
||||
},
|
||||
"README.md": {
|
||||
"bytes": 3696,
|
||||
"sha256": "0e33fe1a85a1ee17ebde57ed7994280dcb95d3a7c4338b78c1f7e0ae9fdcbb7d"
|
||||
},
|
||||
"verify.py": {
|
||||
"bytes": 10908,
|
||||
"sha256": "1e8ba1cc2759fbe76e14636a37925c9dfce9a6f14bf63569acb76498bd8380ad"
|
||||
},
|
||||
"prepare.py": {
|
||||
"bytes": 3572,
|
||||
"sha256": "147a63f26b7131076df52057f1cc263dfefdf99d937b2c2a819217a46c56105f"
|
||||
}
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,56 @@
|
||||
"""Repackage the explicitly closed captures; reads raw evidence, never runs Cargo."""
|
||||
from pathlib import Path
|
||||
import hashlib, json, sys, zipfile
|
||||
|
||||
OUT=Path(__file__).resolve().parent
|
||||
ROOT=Path(sys.argv[1]).resolve() if len(sys.argv)>1 else OUT.parents[3]
|
||||
GATES=ROOT/'.local/transition-gates'
|
||||
ORACLE=ROOT/'.local/transition-native/oracle-falsifiers-1'
|
||||
RUNS=['footer-format-1','footer-windows-build-1','footer-windows-build-2','footer-clippy-1',
|
||||
'lifecycle-producer-build-1','lifecycle-producer-generate-1','lifecycle-producer-build-2',
|
||||
'footer-windows-build-3','footer-format-2','lifecycle-producer-format-1']
|
||||
def pin(b):return {'bytes':len(b),'sha256':hashlib.sha256(b).hexdigest()}
|
||||
def main():
|
||||
payload={};logical={};run_specs={}
|
||||
def add(name,b):
|
||||
# Deduplicate only byte-identical input manifests and source archives.
|
||||
stored='capsules/'+pin(b)['sha256']+Path(name).suffix if Path(name).name in ('inputs.json','inputs.zip') else name
|
||||
if stored in payload:assert payload[stored]==b
|
||||
payload[stored]=b;logical[name]=stored
|
||||
for run in RUNS:
|
||||
result=json.loads((GATES/run/'result.json').read_bytes())
|
||||
assert 'finish' in result and result['inputsUnchanged'] is True,run
|
||||
for f in sorted((GATES/run).iterdir()):
|
||||
assert f.is_file(),f
|
||||
add('cargo/'+run+'/'+f.name,f.read_bytes())
|
||||
helpers=[f for f in GATES.glob('*.py') if pin(f.read_bytes())==result['helper']]
|
||||
assert helpers,run+' helper unavailable'
|
||||
helper='helpers/'+result['helper']['sha256']+'.py'
|
||||
add(helper,helpers[0].read_bytes())
|
||||
run_specs[run]={k:result[k] for k in ('command','head','exitCode','inputCount')}
|
||||
run_specs[run]['helper']=helper
|
||||
for f in sorted(ORACLE.rglob('*')):
|
||||
if f.is_file():add('oracle/'+f.relative_to(ORACLE).as_posix(),f.read_bytes())
|
||||
summary=json.loads((ORACLE/'summary.json').read_bytes())
|
||||
assert summary['status']=='PASS' and summary['trialCount']==40
|
||||
manifest={'schema':1,'captureCheckout':'d0b7c285018a516c6dc34c196d1359a861ddaeb3',
|
||||
'cargo':run_specs,'oracleExpectedTrials':summary['trials'],
|
||||
'oracleOriginalDirectory':str(ORACLE).replace('\\','/'),
|
||||
'logicalFiles':dict(sorted(logical.items())),
|
||||
'members':{n:pin(b) for n,b in sorted(payload.items())}}
|
||||
with zipfile.ZipFile(OUT/'evidence.zip','w',compression=zipfile.ZIP_DEFLATED,compresslevel=9) as z:
|
||||
for name,b in sorted(payload.items()):
|
||||
info=zipfile.ZipInfo(name,date_time=(2026,9,13,0,0,0));info.compress_type=zipfile.ZIP_DEFLATED
|
||||
z.writestr(info,b)
|
||||
(OUT/'manifest.json').write_bytes((json.dumps(manifest,indent=2)+'\n').encode())
|
||||
for name in ('README.md','verify.py','prepare.py'):
|
||||
(OUT/name).write_bytes((OUT/name).read_bytes().replace(b'\r\n',b'\n'))
|
||||
files={name:pin((OUT/name).read_bytes()) for name in ('evidence.zip','manifest.json','README.md','verify.py','prepare.py')}
|
||||
(OUT/'files.json').write_bytes((json.dumps(files,indent=2)+'\n').encode())
|
||||
with zipfile.ZipFile(OUT/'evidence.zip') as z:
|
||||
assert all(z.read(logical[n])==b for n,b in (
|
||||
[('cargo/'+r+'/'+f.name,f.read_bytes()) for r in RUNS for f in sorted((GATES/r).iterdir())]+
|
||||
[('oracle/'+f.relative_to(ORACLE).as_posix(),f.read_bytes()) for f in sorted(ORACLE.rglob('*')) if f.is_file()]))
|
||||
print(json.dumps(files,indent=2))
|
||||
print('Exact raw fidelity:',len(logical),'logical files;',len(payload),'stored members')
|
||||
if __name__=='__main__':main()
|
||||
@@ -0,0 +1,132 @@
|
||||
"""Read-only standard-library verification; no archived source/binary execution."""
|
||||
from pathlib import Path, PurePosixPath
|
||||
from datetime import datetime
|
||||
import hashlib, io, json, zipfile
|
||||
ROOT=Path(__file__).resolve().parent
|
||||
def pin(b):return {'bytes':len(b),'sha256':hashlib.sha256(b).hexdigest()}
|
||||
def need(ok,msg):
|
||||
if not ok:raise ValueError(msg)
|
||||
def safe(n):return not PurePosixPath(n).is_absolute() and '..' not in PurePosixPath(n).parts and '\\' not in n and ':' not in n
|
||||
def members(z,expected):
|
||||
names=z.namelist()
|
||||
need(len(names)==len(set(names)) and set(names)==set(expected),'archive membership')
|
||||
need(all(safe(n) for n in names),'unsafe archive member')
|
||||
def time_order(*values):return list(map(datetime.fromisoformat,values))==sorted(map(datetime.fromisoformat,values))
|
||||
def main():
|
||||
files=json.loads((ROOT/'files.json').read_bytes())
|
||||
need(set(files)=={'evidence.zip','manifest.json','README.md','verify.py','prepare.py'},'outer inventory')
|
||||
for n,p in files.items():need(pin((ROOT/n).read_bytes())==p,'outer hash '+n)
|
||||
m=json.loads((ROOT/'manifest.json').read_bytes())
|
||||
with zipfile.ZipFile(ROOT/'evidence.zip') as z:
|
||||
members(z,m['members'])
|
||||
for n,p in m['members'].items():need(pin(z.read(n))==p,'member hash '+n)
|
||||
logical=m['logicalFiles']
|
||||
need(all(safe(n) for n in logical) and set(logical.values())==set(m['members']),'logical mapping')
|
||||
def raw(n):return z.read(logical[n])
|
||||
def obj(n):return json.loads(raw(n))
|
||||
def directory(prefix):return {n[len(prefix):]:pin(raw(n)) for n in logical if n.startswith(prefix)}
|
||||
pairs=set()
|
||||
for run,spec in m['cargo'].items():
|
||||
p='cargo/'+run+'/'
|
||||
start=obj(p+'started.json');result=obj(p+'result.json');process=obj(p+'process.json')
|
||||
need(all(result[k]==v for k,v in start.items()),run+' start/result')
|
||||
need(result['command']==process['command']==spec['command'],run+' commands')
|
||||
need(result['head']==spec['head'] and result['exitCode']==spec['exitCode'],run+' recorded outcome')
|
||||
need(result['inputsUnchanged'] is True and result['seconds']>=0,run+' source/duration')
|
||||
need(process['pid']>0 and time_order(start['start'],process['observed'],result['finish']),run+' process times')
|
||||
for field,n in [('inputs','inputs.json'),('archive','inputs.zip'),('stdout','stdout.log'),('stderr','stderr.log')]:need(result[field]==pin(raw(p+n)),run+' '+field)
|
||||
need(result['helper']==pin(raw(spec['helper'])),run+' helper')
|
||||
inputs=obj(p+'inputs.json')
|
||||
need(len(inputs)==result['inputCount']==spec['inputCount'],run+' input count')
|
||||
need(sum(n.startswith('crates/vendor/iced_widget/') for n in inputs)==71,run+' vendor inventory')
|
||||
pair=(logical[p+'inputs.json'],logical[p+'inputs.zip'])
|
||||
if pair not in pairs:
|
||||
with zipfile.ZipFile(io.BytesIO(raw(p+'inputs.zip'))) as source:
|
||||
members(source,inputs)
|
||||
for n,expected in inputs.items():need(pin(source.read(n))==expected,run+' source '+n)
|
||||
pairs.add(pair)
|
||||
artifacts=[]
|
||||
for line in raw(p+'stdout.log').decode('utf-8',errors='replace').splitlines():
|
||||
try:item=json.loads(line)
|
||||
except json.JSONDecodeError:continue
|
||||
if isinstance(item,dict) and item.get('reason')=='compiler-artifact' and item.get('executable'):artifacts.append(item)
|
||||
need(artifacts==[a['cargo'] for a in result['artifacts']],run+' artifact relationships')
|
||||
print(run,'exit',result['exitCode'],'inputs',len(inputs),'artifact records',len(artifacts))
|
||||
# Successful build1/generation are not the corrected checker admission.
|
||||
older=obj('cargo/lifecycle-producer-build-1/inputs.json')
|
||||
newer=obj('cargo/lifecycle-producer-build-2/inputs.json')
|
||||
need(older==obj('cargo/lifecycle-producer-generate-1/inputs.json'),'generation uses older producer inputs')
|
||||
source_path='harness/experiments/transition-native-lifecycle-20260913/producer/src/main.rs'
|
||||
need(older[source_path]!=newer[source_path],'distinct producer revisions')
|
||||
final_runs=('footer-windows-build-3','footer-format-2','lifecycle-producer-format-1')
|
||||
final_inputs=obj('cargo/footer-windows-build-3/inputs.json')
|
||||
final_pin='b5a62817a1b5df936b9d759241bdfc4e8283901c1d4b540b9e494995e1544d8d'
|
||||
for run in final_runs:
|
||||
p='cargo/'+run+'/'
|
||||
need(pin(raw(p+'inputs.json'))['sha256']==final_pin and obj(p+'inputs.json')==final_inputs,run+' final source manifest')
|
||||
need(obj(p+'result.json')['exitCode']==0 and len(final_inputs)==715,run+' final successful gate')
|
||||
for run in final_runs[1:]:
|
||||
need(not raw('cargo/'+run+'/stdout.log') and not raw('cargo/'+run+'/stderr.log'),run+' empty format streams')
|
||||
final_build=obj('cargo/footer-windows-build-3/result.json')
|
||||
need(len(final_build['artifacts'])==1 and final_build['artifacts'][0]['cargo']['target']['name']=='arcadia-studio','final GUI artifact')
|
||||
changed={n for n in set(newer)|set(final_inputs) if newer.get(n)!=final_inputs.get(n)}
|
||||
need(changed=={'crates/oa-desktop/src/workflow.rs'},'final wording-only file scope')
|
||||
with zipfile.ZipFile(io.BytesIO(raw('cargo/lifecycle-producer-build-2/inputs.zip'))) as before, zipfile.ZipFile(io.BytesIO(raw('cargo/footer-windows-build-3/inputs.zip'))) as after:
|
||||
path='crates/oa-desktop/src/workflow.rs';a=before.read(path);b=after.read(path)
|
||||
old='{} changes · {} boundary choices · {} conflicts · {} warnings'.encode()
|
||||
new='{} reviewed items · {} boundary choices · {} conflicts · {} warnings'.encode()
|
||||
need(a.count(old)==1 and a.replace(old,new,1)==b,'exact final footer label delta')
|
||||
build=obj('cargo/lifecycle-producer-build-2/result.json')
|
||||
artifact=[a for a in build['artifacts'] if a['cargo']['target']['name']=='transition-native-lifecycle']
|
||||
need(len(artifact)==1,'corrected checker artifact')
|
||||
# Oracle copied build capsule must match the independently included gate.
|
||||
for name in ('result.json','started.json','process.json','inputs.json','inputs.zip','stdout.log','stderr.log'):
|
||||
need(raw('oracle/build-capture/'+name)==raw('cargo/lifecycle-producer-build-2/'+name),'oracle build capsule '+name)
|
||||
need(pin(raw('oracle/build-capture/build-helper.py'))==build['helper'],'oracle build helper')
|
||||
start=obj('oracle/started.json');summary=obj('oracle/summary.json')
|
||||
need(summary['status']=='PASS' and summary['trialCount']==40,'oracle completion')
|
||||
need(all(summary['context'][k]==v for k,v in start.items() if k!='start'),'oracle context')
|
||||
need(start['checker']==artifact[0]['readback'] and start['buildReceipt']==pin(raw('cargo/lifecycle-producer-build-2/result.json')),'oracle checker chain')
|
||||
need(start['runner']==pin(raw('oracle/oracle-falsifiers.py')),'oracle runner pin')
|
||||
for path,p in directory('oracle/producer/').items():need(p==newer['harness/experiments/transition-native-lifecycle-20260913/producer/'+path],'oracle producer source '+path)
|
||||
fixtures=directory('oracle/fixtures/')
|
||||
need(fixtures==start['fixtureInput']==directory('oracle/repeat-generation-a/')==directory('oracle/repeat-generation-b/'),'deterministic repeated fixtures')
|
||||
need(directory('oracle/specimens/')==summary['specimens'],'mutation inventory')
|
||||
rows=summary['trials']
|
||||
need(rows==m['oracleExpectedTrials'] and len(rows)==len({r['label'] for r in rows})==40,'forty distinct trials')
|
||||
actual_labels={n.split('/')[2] for n in logical if n.startswith('oracle/trials/')}
|
||||
need(actual_labels=={r['label'] for r in rows},'exact trial membership')
|
||||
# Original absolute names are mapped to archive members, never opened.
|
||||
original=m['oracleOriginalDirectory'].rstrip('/')+'/'
|
||||
def mapped(path):
|
||||
norm=path.replace('\\','/')
|
||||
need(norm.startswith(original),'unmapped trial input path')
|
||||
rel=norm[len(original):];need(safe(rel),'unsafe trial input mapping')
|
||||
return 'oracle/'+rel
|
||||
for row in rows:
|
||||
label=row['label'];p='oracle/trials/'+label+'/'
|
||||
r=obj(p+'result.json')
|
||||
need(r['label']==label and r['exitCode']==r['expectedExitCode']==row['exitCode']==row['expectedExitCode'],label+' outcome')
|
||||
need(time_order(start['start'],r['start'],r['finish'],summary['finish']),label+' time order')
|
||||
need(r['checkerBefore']==r['checkerAfter']==start['checker'] and r['runner']==start['runner'] and r['buildReceipt']==start['buildReceipt'],label+' provenance')
|
||||
for stream in ('stdout','stderr'):need(r[stream]==pin(raw(p+stream+'.log')),label+' '+stream)
|
||||
need(r['watchedBefore']==r['watchedAfter'],label+' unchanged inputs')
|
||||
for path,expected in r['watchedBefore'].items():need(pin(raw(mapped(path)))==expected,label+' watched input')
|
||||
cmd=r['command'];need(cmd[0]==artifact[0]['cargo']['executable'],label+' executable command')
|
||||
if cmd[1]=='check':
|
||||
need(mapped(cmd[3]) in logical,label+' check entry')
|
||||
opts=cmd[4:];need(len(opts)%2==0,label+' option arity')
|
||||
for flag,path in zip(opts[::2],opts[1::2]):
|
||||
need(flag in ('--identities','--write-identities') and mapped(path) in logical,label+' identity option')
|
||||
if r['exitCode']==0:
|
||||
report=obj(p+'stdout.log')
|
||||
need(report['status']=='PASS' and report['comparison']=='complete Project equality' and report['state']==cmd[2] and report['entry']==cmd[3],label+' positive report')
|
||||
if cmd[2].endswith(('-reopened','-noop')):need(report['trace_identity_pinned'] is True and '--identities' in opts,label+' continuity receipt')
|
||||
if '--write-identities' in opts:need(obj(p+'written-identities.json')==report['new_ids']==obj(mapped(opts[opts.index('--write-identities')+1])),label+' identity write')
|
||||
else:need(cmd[1]=='generate' and len(cmd)==3,label+' generation command')
|
||||
stderr=raw(p+'stderr.log').decode('utf-8',errors='replace')
|
||||
if r['expectedStderrFragment']:need(r['expectedStderrFragment'] in stderr,label+' rejection reason')
|
||||
elif r['exitCode']==0:need(not stderr,label+' successful stderr')
|
||||
print('PASS:',len(m['cargo']),'closed Cargo runs; 40 recorded oracle trials;',len(logical),'logical raw files;',len(m['members']),'stored members;',len(pairs),'source capsule pairs.')
|
||||
print('Offline consistency only: no binaries/source executed; no native lifecycle completion claim.')
|
||||
if __name__=='__main__':main()
|
||||
@@ -0,0 +1,45 @@
|
||||
# Closed native transition lifecycle captures
|
||||
|
||||
Keep this directory beside `../checks/`. Run `python verify.py` with Python 3.10
|
||||
or newer. Verification requires the exact pinned `../checks/evidence.zip` and
|
||||
these native package files, uses only the standard library, and executes no
|
||||
archived helper, application, checker, Cargo command or network request.
|
||||
`prepare.py [CHECKOUT]` rebuilds this archive from the named closed originals.
|
||||
|
||||
Trials lifecycle-2/3 capture OA deletion, recreation, first relink, a separate
|
||||
reopening and no-op check. Trials lifecycle-4/5 capture reverse SA deletion,
|
||||
Undo/Redo, recreation, first relink, reopening and no-op. All four GUI receipts
|
||||
close successfully with empty stdout/stderr and unchanged fixture/binary pins.
|
||||
Trial1 is excluded entirely because it lacked a closing process receipt. This
|
||||
package does not turn that interrupted history into successful evidence.
|
||||
|
||||
Trials 2/3 use footer-windows-build-2; trials 4/5 use footer-windows-build-3 with
|
||||
the refined "reviewed items" footer. Snapshot checks use the corrected producer
|
||||
build2. Exact build receipts, helpers, artifact records and complete source
|
||||
capsules are read from the pinned sibling checks archive; source proof is not
|
||||
self-contained in the native archive alone. Runtime binaries are excluded.
|
||||
|
||||
Every original frame, adjacent metadata, trial file, accepted identity receipt,
|
||||
saved source snapshot and helper is retained without editing. The launch-pinned
|
||||
captureHelper refers to the corresponding capture-lifecycle-N.js source. Each
|
||||
trial's separately recorded capture-function.js is the serialized function used
|
||||
in the persistent UI session; it is different bytes and was not that launch pin.
|
||||
The files document the distinction rather than asserting byte equivalence or
|
||||
authenticated execution. Frame labels/timestamps/application paths record states,
|
||||
not a complete action invocation log or independent PID attestation. Image/window
|
||||
dimensions vary; these captures are not all minimum-window checks.
|
||||
|
||||
The verifier checks exact safe archive membership, hashes, linked build/source
|
||||
capsules, trial/helper/stream relationships, frame membership and timestamps,
|
||||
snapshot helper and corrected-checker provenance, and the recorded complete
|
||||
Project comparison results. Accepted first-relink IDs match later reopen/no-op
|
||||
receipts in each direction, and both later saved source bundles are byte-identical
|
||||
to their first-relink bundle. The producer's full-model result is recorded evidence;
|
||||
the portable verifier does not rerun its Rust comparisons or derive model equality
|
||||
from screenshots. Snapshot receipts record start time but no completion timestamp;
|
||||
only their start can be located in the GUI process interval.
|
||||
|
||||
These finite workflows do not establish exhaustive usability/accessibility,
|
||||
performance, deployment readiness, or behavior on another platform. No credentials,
|
||||
whole user profile, or runtime binaries are packaged. Outer text is UTF-8 LF;
|
||||
archived raw bytes and historical local-path strings remain unchanged.
|
||||
Binary file not shown.
@@ -0,0 +1,22 @@
|
||||
{
|
||||
"evidence.zip": {
|
||||
"bytes": 6888298,
|
||||
"sha256": "b53253fbe9b98db6a0028ed3629426460e7086c8943be0d13ab5ec540e7e4b68"
|
||||
},
|
||||
"manifest.json": {
|
||||
"bytes": 56348,
|
||||
"sha256": "f6bd0bb5e3d29d5e47f1dfb8f1cb7747a3c2e12aa62b6e4e81b68e68ce5ccf95"
|
||||
},
|
||||
"README.md": {
|
||||
"bytes": 3019,
|
||||
"sha256": "01aed34547156bede9c2fea75162c8fecda18203fa575c4cf81e28e440130a95"
|
||||
},
|
||||
"verify.py": {
|
||||
"bytes": 10474,
|
||||
"sha256": "5f93ff8b266c0c7c3e64630f64a522ff5f8f8bd41d7d0a14cc4856c16a529306"
|
||||
},
|
||||
"prepare.py": {
|
||||
"bytes": 2861,
|
||||
"sha256": "c1884b10724cc618b58abd84b93c03623ea9e80dbda4e5a8b866fff499e93fca"
|
||||
}
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,37 @@
|
||||
"""Package only closed lifecycle trials 2-5; never runs a GUI or checker."""
|
||||
from pathlib import Path
|
||||
import hashlib,json,sys,zipfile
|
||||
OUT=Path(__file__).resolve().parent
|
||||
ROOT=Path(sys.argv[1]).resolve() if len(sys.argv)>1 else OUT.parents[3]
|
||||
SRC=ROOT/'.local/transition-native'
|
||||
def pin(b):return {'bytes':len(b),'sha256':hashlib.sha256(b).hexdigest()}
|
||||
def main():
|
||||
payload={};runs={}
|
||||
for number in (2,3,4,5):
|
||||
run=f'lifecycle-{number}';folder=SRC/run
|
||||
result=json.loads((folder/'result.json').read_bytes())
|
||||
assert result['exitCode']==0 and 'finish' in result,run+' not closed successfully'
|
||||
for f in sorted(folder.rglob('*')):
|
||||
if f.is_file():payload['native/'+f.relative_to(SRC).as_posix()]=f.read_bytes()
|
||||
helpers={}
|
||||
for field,name in [('runner',f'run-lifecycle-{number}.py'),('launchCapture',f'capture-lifecycle-{number}.js'),('snapshot',f'check-lifecycle-snapshot-{number}.py')]:
|
||||
helpers[field]='helpers/'+name;payload[helpers[field]]=(SRC/name).read_bytes()
|
||||
captures=[n for n,b in payload.items() if n.startswith('native/'+run+'/') and n.endswith('.json') and isinstance(json.loads(b),dict) and 'images' in json.loads(b)]
|
||||
snapshots={f.name:json.loads((f/'result.json').read_bytes())['command'][2] for f in sorted((folder/'snapshots').iterdir()) if f.is_dir()}
|
||||
runs[run]={'helpers':helpers,'guiBuild':'footer-windows-build-2' if number<4 else 'footer-windows-build-3','captures':captures,'snapshots':snapshots}
|
||||
for f in sorted((SRC/'lifecycle-input-1').rglob('*')):
|
||||
if f.is_file():payload['fixtures/'+f.relative_to(SRC/'lifecycle-input-1').as_posix()]=f.read_bytes()
|
||||
checks=OUT.parent/'checks'
|
||||
payload['references/checks-manifest.json']=(checks/'manifest.json').read_bytes()
|
||||
manifest={'schema':1,'checksEvidence':pin((checks/'evidence.zip').read_bytes()),
|
||||
'checksRelativePath':'../checks/evidence.zip','originalNativeDirectory':str(SRC).replace('\\','/'),
|
||||
'runs':runs,'members':{n:pin(b) for n,b in sorted(payload.items())}}
|
||||
with zipfile.ZipFile(OUT/'evidence.zip','w',compression=zipfile.ZIP_DEFLATED,compresslevel=9) as z:
|
||||
for n,b in sorted(payload.items()):
|
||||
info=zipfile.ZipInfo(n,date_time=(2026,9,22,0,0,0));info.compress_type=zipfile.ZIP_DEFLATED;z.writestr(info,b)
|
||||
(OUT/'manifest.json').write_bytes((json.dumps(manifest,indent=2)+'\n').encode())
|
||||
for n in ('README.md','verify.py','prepare.py'):(OUT/n).write_bytes((OUT/n).read_bytes().replace(b'\r\n',b'\n'))
|
||||
files={n:pin((OUT/n).read_bytes()) for n in ('evidence.zip','manifest.json','README.md','verify.py','prepare.py')}
|
||||
(OUT/'files.json').write_bytes((json.dumps(files,indent=2)+'\n').encode())
|
||||
print(json.dumps(files,indent=2));print('Native raw members:',len(payload))
|
||||
if __name__=='__main__':main()
|
||||
@@ -0,0 +1,115 @@
|
||||
"""Read-only offline verification. Requires the pinned sibling checks/evidence.zip."""
|
||||
from pathlib import Path,PurePosixPath
|
||||
from datetime import datetime
|
||||
import hashlib,io,json,zipfile
|
||||
ROOT=Path(__file__).resolve().parent
|
||||
def pin(b):return {'bytes':len(b),'sha256':hashlib.sha256(b).hexdigest()}
|
||||
def need(ok,msg):
|
||||
if not ok:raise ValueError(msg)
|
||||
def safe(n):return not PurePosixPath(n).is_absolute() and '..' not in PurePosixPath(n).parts and '\\' not in n and ':' not in n
|
||||
def members(z,expected):
|
||||
names=z.namelist();need(len(names)==len(set(names)) and set(names)==set(expected),'exact ZIP membership');need(all(safe(n) for n in names),'safe ZIP members')
|
||||
def ordered(*times):return list(map(datetime.fromisoformat,times))==sorted(map(datetime.fromisoformat,times))
|
||||
def main():
|
||||
files=json.loads((ROOT/'files.json').read_bytes())
|
||||
need(set(files)=={'evidence.zip','manifest.json','README.md','verify.py','prepare.py'},'outer inventory')
|
||||
for n,p in files.items():need(pin((ROOT/n).read_bytes())==p,'outer hash '+n)
|
||||
m=json.loads((ROOT/'manifest.json').read_bytes())
|
||||
need(m['checksRelativePath']=='../checks/evidence.zip','fixed sibling evidence location')
|
||||
checks_bytes=(ROOT/'../checks/evidence.zip').read_bytes()
|
||||
need(pin(checks_bytes)==m['checksEvidence'],'pinned sibling checks archive')
|
||||
with zipfile.ZipFile(ROOT/'evidence.zip') as z,zipfile.ZipFile(io.BytesIO(checks_bytes)) as checks:
|
||||
members(z,m['members'])
|
||||
for n,p in m['members'].items():need(pin(z.read(n))==p,'native member hash '+n)
|
||||
def raw(n):return z.read(n)
|
||||
def obj(n):return json.loads(raw(n))
|
||||
def directory(prefix):return {n[len(prefix):]:pin(raw(n)) for n in m['members'] if n.startswith(prefix)}
|
||||
cm=obj('references/checks-manifest.json');members(checks,cm['members'])
|
||||
def craw(logical):
|
||||
n=cm['logicalFiles'][logical];b=checks.read(n);need(pin(b)==cm['members'][n],'checks member '+n);return b
|
||||
def cobj(n):return json.loads(craw(n))
|
||||
# Source capsules and receipt helper chains are independently checked here.
|
||||
builds={};artifacts={};pairs=set()
|
||||
for run in ('footer-windows-build-2','footer-windows-build-3','lifecycle-producer-build-2'):
|
||||
p='cargo/'+run+'/';r=cobj(p+'result.json');start=cobj(p+'started.json');process=cobj(p+'process.json')
|
||||
need(all(r[k]==v for k,v in start.items()) and r['exitCode']==0 and r['inputsUnchanged'] is True,run+' build closure')
|
||||
need(r['command']==process['command'] and ordered(r['start'],process['observed'],r['finish']),run+' build process')
|
||||
for field,n in [('inputs','inputs.json'),('archive','inputs.zip'),('stdout','stdout.log'),('stderr','stderr.log')]:need(r[field]==pin(craw(p+n)),run+' receipt '+field)
|
||||
need(r['helper']==pin(craw(cm['cargo'][run]['helper'])),run+' build helper')
|
||||
inputs=cobj(p+'inputs.json');need(len(inputs)==r['inputCount'] and sum(n.startswith('crates/vendor/iced_widget/') for n in inputs)==71,run+' input inventory')
|
||||
pair=(r['inputs']['sha256'],r['archive']['sha256'])
|
||||
if pair not in pairs:
|
||||
with zipfile.ZipFile(io.BytesIO(craw(p+'inputs.zip'))) as source:
|
||||
members(source,inputs)
|
||||
for n,expected in inputs.items():need(pin(source.read(n))==expected,run+' source '+n)
|
||||
pairs.add(pair)
|
||||
records=[]
|
||||
for line in craw(p+'stdout.log').decode('utf-8',errors='replace').splitlines():
|
||||
try:item=json.loads(line)
|
||||
except json.JSONDecodeError:continue
|
||||
if isinstance(item,dict) and item.get('reason')=='compiler-artifact' and item.get('executable'):records.append(item)
|
||||
need(records==[a['cargo'] for a in r['artifacts']],run+' artifact records')
|
||||
expected='transition-native-lifecycle' if run.startswith('lifecycle') else 'arcadia-studio'
|
||||
selected=[a for a in r['artifacts'] if a['cargo']['target']['name']==expected];need(len(selected)==1,run+' unique artifact')
|
||||
builds[run]=r;artifacts[run]=selected[0]
|
||||
original=m['originalNativeDirectory'].rstrip('/')+'/'
|
||||
def mapped(path):
|
||||
norm=path.replace('\\','/');need(norm.startswith(original),'snapshot path outside captured native root')
|
||||
rel=norm[len(original):];need(safe(rel),'unsafe snapshot path');return 'native/'+rel
|
||||
fixtures=directory('fixtures/');snapshots={};reports={};capture_count=image_count=0
|
||||
need(set(m['runs'])=={'lifecycle-2','lifecycle-3','lifecycle-4','lifecycle-5'},'closed trial membership')
|
||||
for run,spec in m['runs'].items():
|
||||
p='native/'+run+'/';r=obj(p+'result.json');start=obj(p+'started.json');process=obj(p+'process.json')
|
||||
build=builds[spec['guiBuild']];artifact=artifacts[spec['guiBuild']]
|
||||
need(all(r[k]==v for k,v in start.items()),run+' start/result relationship')
|
||||
need(r['exitCode']==0 and r['inputUnchanged'] is True and r['binary']==r['binaryAfter']==artifact['readback'],run+' native closure and binary pin')
|
||||
need(r['buildReceipt']==pin(craw('cargo/'+spec['guiBuild']+'/result.json')) and r['buildHead']==build['head'],run+' GUI build relationship')
|
||||
need(r['command']==[artifact['cargo']['executable']] and process['pid']>0,run+' native command/PID')
|
||||
need(ordered(build['finish'],r['start'],process['observed'],r['finish']),run+' native times')
|
||||
need(r['helper']==pin(raw(spec['helpers']['runner'])) and r['captureHelper']==pin(raw(spec['helpers']['launchCapture'])),run+' launch helper pins')
|
||||
need(p+'capture-function.js' in m['members'],run+' actual capture-function record')
|
||||
need(r['inputs']==pin(raw(p+'inputs.json')) and obj(p+'inputs.json')==fixtures,run+' unchanged fixture bytes')
|
||||
for stream in ('stdout','stderr'):need(not raw(p+stream+'.log') and r[stream]==pin(raw(p+stream+'.log')),run+' empty '+stream)
|
||||
images=[]
|
||||
actual_captures=[n for n in m['members'] if n.startswith(p) and n.endswith('.json') and isinstance(obj(n),dict) and 'images' in obj(n)]
|
||||
need(set(actual_captures)==set(spec['captures']) and len(actual_captures)==len(spec['captures']),run+' exact metadata coverage')
|
||||
for n in spec['captures']:
|
||||
frame=obj(n);need(ordered(r['start'],frame['time'],r['finish']),n+' frame timestamp')
|
||||
need(frame['window']['app'].replace('\\','/')=='process:'+r['command'][0].replace('\\','/'),n+' window executable path')
|
||||
for image in frame['images']:
|
||||
name=p+image['file'];b=raw(name)
|
||||
need(image['width']>0 and image['height']>0 and (b.startswith(b'\xff\xd8') or b.startswith(b'\x89PNG')),n+' image record')
|
||||
images.append(name)
|
||||
actual_images={n for n in m['members'] if n.startswith(p) and n.endswith(('.jpg','.png'))}
|
||||
need(len(images)==len(set(images)) and set(images)==actual_images,run+' all original images')
|
||||
capture_count+=len(spec['captures']);image_count+=len(images)
|
||||
actual_snapshots={n.split('/')[3] for n in m['members'] if n.startswith(p+'snapshots/')}
|
||||
need(actual_snapshots==set(spec['snapshots']),run+' snapshot inventory')
|
||||
for label,state in spec['snapshots'].items():
|
||||
q=p+'snapshots/'+label+'/';s=obj(q+'result.json');ss=obj(q+'started.json');report=obj(q+'stdout.log')
|
||||
checker=artifacts['lifecycle-producer-build-2']
|
||||
need(all(s[k]==v for k,v in ss.items()) and s['exitCode']==0 and s['binaryUnchanged'] is True and s['workingSourceUnchanged'] is True,label+' snapshot closure')
|
||||
need(ordered(r['start'],s['time'],r['finish']),label+' snapshot start during trial')
|
||||
need(s['binary']==checker['readback'] and s['buildReceipt']==pin(craw('cargo/lifecycle-producer-build-2/result.json')),label+' corrected checker build')
|
||||
need(s['helper']==pin(raw(spec['helpers']['snapshot'])),label+' snapshot helper pin')
|
||||
need(s['source']==directory(q+'source/'),label+' exact saved source bytes')
|
||||
for stream in ('stdout','stderr'):need(s[stream]==pin(raw(q+stream+'.log')),label+' '+stream)
|
||||
need(not raw(q+'stderr.log'),label+' empty snapshot stderr')
|
||||
cmd=s['command'];need(cmd[:3]==[checker['cargo']['executable'],'check',state] and mapped(cmd[3])==q+'source/original.arcadia',label+' checker command')
|
||||
need(report['status']=='PASS' and report['comparison']=='complete Project equality' and report['state']==state and report['entry']==cmd[3],label+' accepted model oracle')
|
||||
opts=cmd[4:];need(len(opts)%2==0,label+' option pairs')
|
||||
for flag,path in zip(opts[::2],opts[1::2]):
|
||||
need(flag in ('--identities','--write-identities'),label+' identity-only options')
|
||||
need(obj(mapped(path))==report['new_ids'],label+' exact identity receipt')
|
||||
if state.endswith(('-reopened','-noop')):need('--identities' in opts and report['trace_identity_pinned'] is True,label+' continuity identity pin')
|
||||
if state.endswith('-relinked'):need('--write-identities' in opts and mapped(report['identity_receipt'])==mapped(opts[opts.index('--write-identities')+1]),label+' initial accepted identities')
|
||||
snapshots[(run,label)]=s['source'];reports[(run,label)]=report
|
||||
print(run,'captures',len(spec['captures']),'images',len(images),'accepted snapshots',len(spec['snapshots']))
|
||||
for side,first,later in [('oa','lifecycle-2','lifecycle-3'),('sa','lifecycle-4','lifecycle-5')]:
|
||||
base=snapshots[(first,side+'-relinked')];ids=reports[(first,side+'-relinked')]['new_ids']
|
||||
for phase in ('reopened','noop'):
|
||||
key=(later,side+'-'+phase)
|
||||
need(snapshots[key]==base and reports[key]['new_ids']==ids,side+' saved bytes and identity continuity '+phase)
|
||||
print('PASS: 4 closed trials;',len(snapshots),'accepted full-model receipts;',capture_count,'capture records;',image_count,'original images;',len(pairs),'linked source capsules.')
|
||||
print('No UI history authentication or checker re-execution; snapshot receipts lack completion timestamps. Frame dimensions vary.')
|
||||
if __name__=='__main__':main()
|
||||
@@ -0,0 +1,102 @@
|
||||
# Independent bounded transition review
|
||||
|
||||
Reviewer: `transition_picker_review`, 2026-09-22. This review inspected code,
|
||||
original native screenshots and closed-run/snapshot receipts. The reviewer did
|
||||
not operate the application, build it, rerun the checker, or alter the specimens.
|
||||
Only this review note was written. The parent owns execution and publication.
|
||||
|
||||
The reviewed production baseline is `d0b7c285018a516c6dc34c196d1359a861ddaeb3`
|
||||
plus the captured, initially uncommitted footer changes in
|
||||
`crates/oa-desktop/src/workflow.rs`. HEAD alone does not identify those changes.
|
||||
Lifecycle trials 2 and 3 identify GUI SHA-256
|
||||
`3813e79d84cf4a83f2290639b93c3daaa8de427b0268b288f87e4f535fa1aea2`;
|
||||
trials 4 and 5 identify final-footer GUI SHA-256
|
||||
`f0bb021e350bcc55435d29e3daf6a4b3dd7389620df99f5d169f9cb669fd53d0`.
|
||||
Exact build/source inventories belong to the accompanying evidence packages.
|
||||
|
||||
## Code and evaluator findings
|
||||
|
||||
The picker shares the core layer/category/kind predicate, preserves selected
|
||||
identity independently of search results, and keeps query state outside the
|
||||
request and recovery schema. Query events return before replanning. Compatibility
|
||||
does not imply that dependencies or the complete transition plan are valid.
|
||||
Distinct traced IDs, rather than trace-record count, determine inferred reuse;
|
||||
ambiguous traces remain an explicit decision. Manual tracing remains separate.
|
||||
|
||||
The desktop tests compare the whole request, plan and recovery snapshot during
|
||||
search, including generated preview identities. The initial second-query test
|
||||
used a target ID rather than a second rendered source card; native corrected
|
||||
picker frame 17 separately demonstrates two visible source cards with independent
|
||||
queries. These checks do not prove keyboard access or every focus transition.
|
||||
|
||||
The core lifecycle tests and [native producer/checker](producer/src/main.rs)
|
||||
construct expected deletions from fixture-owned identities, not a deletion
|
||||
impact helper or observed deletion result. Expected recreation adds only the
|
||||
replacement root. Expected relinking adds a directly constructed refinement
|
||||
trace and reviewed baseline while preserving target prose, bodies, ownership,
|
||||
geometry, routes and artifacts. Whole-Project equality detects incidental changes.
|
||||
First-relink admission may observe the fresh trace UUID; reopen/no-op admission
|
||||
requires the previously written identity receipt and rejects replacement IDs.
|
||||
Source loading rejects compiler-inserted identities. Round trips compare the
|
||||
decoded model; they do not promise original comments or formatting preservation.
|
||||
No blocking conservation-oracle defect was found in the inspected paths.
|
||||
|
||||
## Native observations
|
||||
|
||||
Trial/frame labels below refer to original `*-0.jpg` files and matching JSON
|
||||
under `.local/transition-native/` at capture time. Receipt evidence is distinct
|
||||
from what screenshots visibly establish. Capture helpers and checksums provide
|
||||
provenance context, not authenticated historical action logs.
|
||||
|
||||
| Originals inspected | Visible observation |
|
||||
| --- | --- |
|
||||
| `picker-baseline-1` 10–11 | Activity menu includes incompatible ports; selecting one yields a kind conflict. |
|
||||
| `picker-corrected-2` 06–07, 10–12, 15, 17–19, 22–23 | Six compatible Functions are sorted; alias filtering preserves current selection until explicit selection; zero results preserve the selected target; two cards retain independent queries; field review remains required; clearing search retains Keep target; Cancel returns to the saved-status view. |
|
||||
| `lifecycle-2` 05, 11–12 | OA deletion is visible; a replacement root declaration is entered and applied through Textual model. |
|
||||
| `lifecycle-2` 17, 21, 27–30 | Recreation initially offers Create rather than silently relinking. Selecting the retained SA target requires Name and Description decisions. Sticky guidance advances from Name/two conflicts to Description/one, then clears after both Keep target decisions. Seven warnings do not block Apply. |
|
||||
| `lifecycle-2` 31–32 | Apply displays the retained SA root, children and exchange; subsequent frame reports all changes saved. |
|
||||
| `lifecycle-3` 02, 06–08 | A fresh process opens the saved model; unchanged transition infers reuse without field conflicts; Apply and save return to the retained SA design. |
|
||||
| `lifecycle-4` 12, 14, 16, 22–23 | SA deletion, Undo restoration, Redo deletion and replacement-root entry/application are visible. |
|
||||
| `lifecycle-4` 28, 30–38 | Reverse transition defaults to Create; alias search offers the retained OA Activity; explicit selection requires both prose decisions. Persistent guidance follows two/one/zero conflicts. Apply displays the retained OA decomposition and save reports completion. The final `reviewed items` wording is visible. |
|
||||
| `lifecycle-5` 02, 07–09 | A fresh process opens the retained OA design; unchanged SA-to-OA review infers the accepted counterpart without conflicts; Apply and save retain the visible OA design. |
|
||||
|
||||
Closed GUI receipts for lifecycle trials 2–5 record exit 0, empty stdout/stderr,
|
||||
and unchanged executable hashes. Snapshot receipts record exit 0 for OA deletion,
|
||||
recreation and relink; OA reopen/no-op; and SA deletion, original-state Undo,
|
||||
deletion-state Redo, recreation, relink and subsequent reopen/no-op. Checks operate on saved source copies
|
||||
and report the working source unchanged. SA Redo's source hash equals deletion's.
|
||||
|
||||
OA relink, reopen and no-op snapshots have identical hashes for both
|
||||
`original.arcadia` and `traceability.arcadia`. Reopen/no-op use source
|
||||
`8eb83649-2810-58e1-94cd-71e25422658e` and accepted trace
|
||||
`856eb758-33fc-4ebf-9a87-22ad40a89952`. SA first relink establishes source
|
||||
`074eab66-7092-5e7e-8b83-b31b95afdbcb` and trace
|
||||
`acc7969b-08c3-43f4-a31d-9140ca873ffe`. Trial 5 passes that exact first-relink
|
||||
receipt to both saved-source checks. Its reopen and no-op source hashes match the
|
||||
first relink for both files (`original.arcadia`: `e75b4cb55e5b205bd5784a1f35573931f1e64491b646f4ffb241f90288447b5d`;
|
||||
`traceability.arcadia`: `5046e2d2731c5d8fcbcb7fe2deede0677ec61de03b85890852fa20f778c57552`).
|
||||
This supports conservation and persisted identity across these finite reopen and
|
||||
unchanged-review workflows in both directions. No new blocking finding arose
|
||||
from the reviewed lifecycle evidence.
|
||||
|
||||
## Feedback correction and remaining limits
|
||||
|
||||
Corrected picker frames 12/15 initially left Apply disabled while its explanation
|
||||
was below the fold. The reviewed footer change moves counts outside the scrolling
|
||||
body and derives guidance from the same conflicts and destination validity used
|
||||
by Apply. It adds no persisted state and weakens no validation. Generic guidance
|
||||
was corrected from “adjust mapping or scope” because operation-omission rationale
|
||||
and component-port decisions require other controls. Final native trial 4 shows
|
||||
the Name/Description guidance and `reviewed items` count. Destination, boundary
|
||||
and general-conflict guidance were checked against code, not exercised in these
|
||||
native lifecycle frames.
|
||||
|
||||
All inspected originals are 1442×932. Resize attempts did not establish a smaller
|
||||
viewport, regardless of frame filenames; minimum-window behavior is unverified.
|
||||
The inherited picker is pointer-dependent. This review does not establish full
|
||||
keyboard/screen-reader accessibility, packaged launch, other platforms, arbitrary
|
||||
models, implicit creation applied natively, or relink Undo/Redo. Recreated roots
|
||||
were authored through Textual model and intentionally have no new placement;
|
||||
the empty source canvas is consistent with that fixture. Screenshots alone do
|
||||
not establish complete model conservation: the independent saved-source comparisons
|
||||
provide that bounded evidence.
|
||||
@@ -52,4 +52,83 @@ The corrected GUI trial ran 2026-09-13 15:22:22–15:27:13 UTC on `d46854c`. It
|
||||
|
||||
Independent read-only reviews found no blocking defect in the shared predicate, selection/query handling, recovery behavior, or lifecycle conservation evaluator. Two refinements were incorporated: empty-query feedback no longer asks to clear an already empty search, and query-isolation tests use two actual OA source cards. A separate review of original native frames confirmed the bounded observations above without a blocking visual regression. Remaining friction: disabled Apply can have its required field-review explanation below the fold, requiring a scroll. Keyboard-only picker operation and broader transition-dialog feedback remain incomplete.
|
||||
|
||||
The native trial establishes this finite Windows selection/review/cancel workflow. It does not establish native endpoint deletion/recreation/relink, saving an applied transition, packaged launch, or whole-project equality from screenshots. Those lifecycle operations have the core evidence above; the broader C1 native acceptance and integration/release decision remain open.
|
||||
That picker-only trial establishes the finite Windows selection/review/cancel workflow. The applied lifecycle follow-up below supplies separate native and saved-model evidence; neither trial is a packaged-release acceptance claim.
|
||||
|
||||
## Applied native lifecycle follow-up, 2026-09-22
|
||||
|
||||
The original disabled-Apply observation led to a small presentation correction:
|
||||
transition counts and the next blocking decision now remain beside Apply, outside
|
||||
the scrolling mappings. Guidance identifies destination, boundary-kind, field or
|
||||
other conflict decisions; it does not weaken Apply validation. The count says
|
||||
`reviewed items` because a reused item need not change the model. No new persisted
|
||||
state or transition semantics are introduced.
|
||||
|
||||
The [isolated producer/checker](experiments/transition-native-lifecycle-20260913/producer/README.md)
|
||||
defines a deterministic two-layer fixture and independently constructs expected
|
||||
deletion, recreation and relink states. It compares complete Projects, exact
|
||||
engineering-warning multisets, JSON/source round trips and persisted identities.
|
||||
An independent critic exposed two potential false passes: learning replacement
|
||||
identities again after reopening, and accepting identities inserted only into
|
||||
compiler memory. Both were corrected. First relink records accepted identities;
|
||||
reopen/reapply require that unchanged receipt. Forty synthetic falsification
|
||||
trials include deliberate prose/geometry corruption, identity drift, omitted
|
||||
persisted IDs and receipt misuse. They are checker evidence, not GUI actions.
|
||||
|
||||
The [closed Cargo and oracle package](experiments/transition-native-lifecycle-20260913/checks/README.md)
|
||||
retains ten successful invocations, including strict all-target Clippy, final
|
||||
desktop build and both format checks. Earlier checker build1 incorrectly rejected
|
||||
the fixture's expected engineering warnings during native admission; corrected
|
||||
build2 accepts only the exact expected diagnostic multiset. The original failed
|
||||
admission and an interrupted first GUI session were not counted as successful
|
||||
native trials. The final build differs from the linted footer only in the
|
||||
`changes` to `reviewed items` literal; the package verifies that exact delta.
|
||||
|
||||
Four fresh, closed Windows sessions provide the
|
||||
[native evidence](experiments/transition-native-lifecycle-20260913/native/README.md):
|
||||
|
||||
| Flow | Native operations and saved-source acceptance |
|
||||
| --- | --- |
|
||||
| OA source, trials 2–3 | Delete OA root; recreate only that root through Textual model; search `@sa_response`; explicitly select the retained SA function; keep target Name and Description; apply/save; restart/open/save; unchanged review/apply/save. Five full-Project checks pass. |
|
||||
| SA source, trials 4–5 | Delete SA root; Undo/save restores the complete original; Redo/save repeats the deletion; recreate only that root through Textual model; search `@oa_response`; explicitly select the retained OA activity; keep both prose fields; apply/save; restart/open/save; unchanged review/apply/save. Seven full-Project checks pass. |
|
||||
|
||||
The retained opposite-layer design includes children, internal exchanges, typed
|
||||
SA ports, shared contracts, authored diagram geometry/routes and artifact links.
|
||||
Recreated roots intentionally have no children or placement. The relink adds one
|
||||
trace and its reviewed baseline; it does not recreate deleted design work.
|
||||
For each direction, both saved source files remain byte-identical through relink,
|
||||
restart/save and unchanged reapply/save, with the first accepted source/trace IDs
|
||||
pinned. This is evidence about these concrete files, not a general promise to
|
||||
preserve arbitrary authored comments or whitespace.
|
||||
|
||||
All four processes exited zero with empty stdout/stderr and unchanged executable
|
||||
hashes. Trials 2–3 used GUI SHA256
|
||||
`3813e79d84cf4a83f2290639b93c3daaa8de427b0268b288f87e4f535fa1aea2`;
|
||||
trials 4–5 used the final 117,090,816-byte GUI SHA256
|
||||
`f0bb021e350bcc55435d29e3daf6a4b3dd7389620df99f5d169f9cb669fd53d0`.
|
||||
Build receipts record HEAD `d0b7c28` plus captured uncommitted input bytes;
|
||||
the 715-file final inventory, rather than that old HEAD alone, identifies the
|
||||
tested source. All 715 captured inputs were checked against exact staged Git
|
||||
blobs before source commit `05e4c83054dc8b2f156e50e7731a50ac19ae8767`.
|
||||
|
||||
The [independent review](experiments/transition-native-lifecycle-20260913/review.md)
|
||||
separates original-frame observations from saved-model comparisons and found no
|
||||
new blocking defect in this bounded workflow. Native frames show the persistent
|
||||
Name/Description instruction advancing with two, one and zero conflicts, while
|
||||
expected modeling warnings remain nonblocking. Destination, boundary and generic
|
||||
conflict guidance were reviewed in code but not exercised by these native trials.
|
||||
All main-window captures are 1442×932; attempted resizing did not establish
|
||||
minimum-window behavior. Keyboard/screen-reader accessibility, native implicit
|
||||
creation, relink Undo/Redo, arbitrary models and packaged launch remain outside
|
||||
this acceptance. Linux checks remain paused by the user. Integration and release
|
||||
status belong to issue 28; this is branch-specific evidence.
|
||||
|
||||
Run the read-only evidence verifiers with Python 3.10+:
|
||||
|
||||
```text
|
||||
python harness/experiments/transition-native-lifecycle-20260913/checks/verify.py
|
||||
python harness/experiments/transition-native-lifecycle-20260913/native/verify.py
|
||||
```
|
||||
|
||||
These verify archived bytes and recorded relationships without executing archived
|
||||
code or authenticating historical interaction. The producer README gives the
|
||||
commands and native steps to reproduce behavior independently.
|
||||
|
||||
Reference in New Issue
Block a user