Server measures every classified APDU (A4 select, B0 read binary, B2 read
record) from command to response: _collect_apdu_times() enables the tracer
(reattaching it if pySim nulled it) and /api/select + /api/read return
'apdu_times': [{type, ms}]. Collection is safe: handlers hold _CARD_LOCK.
Snapshots store per-file {select_ms, read_ms}, a ms value per record and
snapshot-level stats {select, read_binary, read_record}: {min, max, avg,
count} plus total_ms (wall time of the scan). The snapshot view gets a
summary under the title (files/records counts, scan time, min/avg/max per
command type) and shows select/read per file and read time per record.
Timings are display-only: checks, snapshots comparison and imports ignore
them (old snapshots simply show 'No timing data').
Tests: Python classifier/collection (tests/test_apdu_timing.py) and
frontend stats/accumulator/format/build-file/summary. SW cache v99 ->
v100; help, README and docs/api.md updated.
'Profile from card' and 'New snapshot' now focus the name field when the
dialog opens, and Enter in that field starts the scan
(profilerScanNameKeydown, ignored while the Scan button is disabled).
Tests for the key handler and the input wiring; SW cache v98 -> v99.
When a navigation fetch failed and the cache had no entry for the exact
URL (e.g. '/' while only 'index.html' is precached, common while the
local server restarts), the fetch handler resolved respondWith() to
undefined, producing 'TypeError: Failed to convert value to Response'.
The navigate fallback now chains caches.match(request) -> cached
index.html -> an explicit 503 offline Response. sw.js requests are
network-only with the same offline Response (previously they fell into
the navigate branch), and a new unit test loads sw.js in a VM sandbox
covering navigate/cache/asset/api paths. SW cache v97 -> v98.
After the gray text was raised to #cbd5e1 in dark mode, elements using
dark:text-slate-300 (inactive tabs, labels, list text) matched the muted
level. Map dark:text-slate-300 to #e2e8f0 (slate-200, the body's dark
text color) so normal text stays one step brighter than muted gray.
SW cache v96 -> v97.
Elements using bare text-gray-300/400/500/600 (proactive log ids and
timestamps, expand markers, event codes, ...) kept the light-theme gray
values in dark mode, which made them hard to read on slate-800/900.
- dark .text-gray-*:not([class*='dark:text-']) rules map them to the same
level as the other dark muted text (slate-300 #cbd5e1, slate-200/100 for
gray-700/800)
- dark:text-gray-400/600 and dark:text-slate-500 normalized to #cbd5e1
- dark:border-slate-600/700/800 (+ /50) brightened one step
- light theme unchanged
- the whole contrast block moved to src/contrast.css, appended by the
npm build scripts after Tailwind so it stays reproducible; SW cache
v95 -> v96
Wrap each block of the phone row in a bordered card (border, rounded,
p-3) with items-start and gap-4 so they read as distinct panels and wrap
cleanly. SW cache v94 -> v95.
- Phone pill: STK menu, STATUS and polling and the subscribed-events list
in one row, proactive command log full-width below (room for more
elements)
- TR Config pill: response data injected into TERMINAL RESPONSEs;
currently the PROVIDE LOCAL INFORMATION dictionary, structured as
heading + body blocks for future proactive-command responses
- phoneSwitchSubtab() mirrors the other sub-tab switchers and sets help
anchors stk-menu / pli-dict; switchTab('phone') always opens Phone
- refreshDynamicI18n renders only the visible phone panel
- help EN/RU section 6 regrouped (6.4 STATUS polling under Phone, 6.5
TR Config response data), READMEs and AGENTS.md updated
- structural and behavioral tests (html.test.js, phone_tabs.test.js);
SW cache v93 -> v94
- C-APDU tab renamed Remote APDU (label untranslated EN/RU); Response
parser moved from a top-level tab to a pill under it
- Profiler moved from the Card reader sub-tabs to a top-level tab
- Proactive UICC moved to a top-level tab and renamed Phone simulator
- Card reader keeps File manager, Custom files, pySim command line,
Raw APDU
- modals (event send, profiler scan/snapshot, STK menu) moved outside
the tab containers so they can open from any tab
- STK overlay disables the top-level tabs while waiting for user input
- OTA PoR jump now goes to Remote APDU > Response parser
- help EN/RU restructured (2.8 Response parser, 5 Profiler, 6 Phone
simulator, subsequent sections renumbered, anchors kept); READMEs and
AGENTS.md updated
- html.test.js asserts the new tab/pill structure; SW cache v92 -> v93
- _apply_equipped_card() centralizes the post-equip refresh + TERMINAL
PROFILE (shared by the /api/command equip branch and auto-equip)
- server tracks card_session (bumped on equip and disconnect) and
equipping; /api/status exposes connected, card_present, card_session,
equipping, auto_equip and is exempt from _CARD_LOCK (pure cached state)
- auto-equip is on by default (--no-auto-equip; off with --no-card-init):
the presence observer spawns a one-shot worker after insertion, which
runs equip under _CARD_LOCK and applies the same refresh; the monitor
starts after the startup init so pyscard's initial 'already present'
event does not re-equip a fresh session
- UI: /api/status polls every 2s (other views stay at 5s); when
card_session changes it runs pysimResetCardData() (STK overlay, file
tree, events, proactive log, PLI, status) — the same reset as a manual
Equip; messages: initializing / press Equip / no card
Tests for the observer and auto-equip rules, session bumps,
_apply_equipped_card, and the UI state machine. SW cache v91 -> v92.
The UI only noticed a removed card when some user action ran a real card
command (e.g. Check status); /api/status is a cached-state read that kept
returning the old card, and _handle_card_disconnect() did not clear
app.card/rs.
- start_card_monitor() registers a pyscard CardObserver for our reader;
it only polls SCardGetStatusChange (no APDU, no connection, no extra
process), and on removal sets server.card_present=False and calls
_handle_card_disconnect() under _CARD_LOCK
- /api/status now exposes connected (session usable) and card_present
(physically inserted) and masks card/profile/atr/selection when not
connected; _CARD_CONNECTED is initialized from card presence instead of
being unconditionally True
- the 2s UI poll includes /api/status; on disconnect it switches to the
existing 'No card detected. Insert card and click Equip' state, or the
new 'Card inserted — press Equip' hint when the card is back; the old
_hadData heuristic is gone
Tests for the observer (filtering, removal, insertion) and the UI state
transitions. SW cache v90 -> v91.
Every FETCHed proactive command must be answered, otherwise the card is
left in an unfinished session and stops issuing commands (e.g. it will
not deliver a PoR for SEND SM). The menu handlers now share
_menu_send_response, and a server-side watchdog (_arm_menu_timeout /
_menu_timeout_fire, --menu-timeout, default 60s, 0 disables) sends the
timeout result (0x12) when the user never answers. The timer is armed
while a command is pending and cancelled on any response, equip, rescue
and card disconnect.
Also fixes docs/api.md, which had back (0x11) and timeout (0x12) codes
swapped. Tests for arm/cancel/clamping and the flat timeout TR. SW cache
v89 -> v90.
pysim's init_card() resets the card once per profile candidate in
CardProfile.pick(), once in RuntimeState.__init__ and again in
PysimApp.equip(); on typical readers each reconnect costs ~1.3s and a
normal init/equip does 7-8 of them (measured: equip 9.85s, startup
~14.8s).
fastinit.py mirrors pySim.app.init_card() with the resets removed:
pick_profile_no_reset() runs all profile probes back-to-back on one
connection, FastRuntimeState.reset() is a software reset (select MF,
clear selected_adf/scp, ATR from the transport) and the equip/reset
commands are routed through do_equip_fast (one reconnect via
wait_for_card) and do_reset_fast (always a physical reset). Enabled
with --fast-init; stock behavior remains the default.
Tests for the reset-free pick, the soft reset and the explicit reset
paths. Docs updated; SW cache v88 -> v89.
New Compare snapshots button on the Card snapshots tab opens a dialog
with Master snapshot / Snapshot to check selects and the two mask
options (Match first 4 bytes for EF.IMSI / EF.ICCID, checked by
default). The comparison runs like a profile check where the master
snapshot takes the place of the profile: every file must match exactly
(exact FCI, contents), except the first 4 bytes of masked EF.IMSI/EF.ICCID.
profilerRulesFromSnapshot() synthesizes exact-match rules from the
master; profilerSnapshotSource() of the checked snapshot is the data
source; files present only in the checked snapshot are appended as
failed 'extra file' results via profilerExtraFileResults() and the new
optional extraResults param of profilerRunProfile(). Results reuse the
existing view, summary and Only mismatches filter; Back to list returns
to the snapshots tab. Tests for rule synthesis, masking, comparison and
extra files. SW cache v87 -> v88.
The results view header (live card and snapshot checks) gets an Only
mismatches checkbox that hides all passing files and keeps only failures
and errors; when everything passes under the filter, a 'No mismatches'
note is shown instead. The pass/fail/error summary always reflects all
results. Helps with large profiles on cards where most files match.
SW cache v86 -> v87.
Each profile row gets a Check card snapshot button next to Check card.
It opens a picker listing the stored snapshots (name, ICCID, date, file
count) and runs the profile rules offline against the selected snapshot,
showing the usual report titled 'profile — snapshot'.
profilerRunRule() accepts an optional data source: profilerLiveSource()
(default) wraps /api/select + /api/read, profilerSnapshotSource() serves
select/read from the snapshot files. Contents missing from the snapshot
are reported as unverifiable errors ('Content not captured in snapshot')
rather than mismatches. The shared check loop moved into
profilerRunProfile().
Tests for the snapshot source and snapshot-based checks; help docs
updated; SW cache v84 -> v85.
translatePage() only rewrites static [data-l10n] elements, so labels built
at render time (Check card/Edit/Export/Delete, Open, Remove, Send, ...)
stayed in the old language. toggleLang() now calls refreshDynamicI18n(),
which re-renders only the visible dynamic views:
- profiler sub-tab: current view (list, snapshot list, editor, results via
the new profilerRenderResultsView with profilerResults stored, snapshot
via profilerRenderSnapshotData which leaves the name input alone)
- open scan modal: profilerScanRefreshOptions re-translates the mask labels
without touching checkbox state
- cards table (cardsRender), proactive events/log/PLI (async, only when
the sub-tab is visible)
Also removed the duplicated translatePage() call in toggleLang.
New test for profilerScanRefreshOptions. SW cache v83 -> v84.
Rectangular top-rounded tabs with a bottom border, matching the app's
top-level tab bar (incl. dark:bg-blue-500/dark:text-white active state,
toggled by profilerListSwitch). SW cache v82 -> v83.
The list view now has a centered pill row (Profiles | Card snapshots) and
shows one list at a time; the last active tab persists while the app is
open. Snapshot flows (scan, import) switch to the snapshots tab, profile
import switches to profiles. Help docs reworded (tabs). New DOM test for
profilerListSwitch. SW cache v81 -> v82.
New read-only capture entity stored in localStorage 'otaman_snapshots':
- profiler list view is now two labelled columns: Profiles and Card
snapshots; each snapshot row shows name, decoded ICCID, date, file
count and Open / Export / Delete buttons
- New snapshot reuses the scan modal in a snapshot mode (name only, no
ignore/mask/FCP options) via profilerScanCard(..., 'snapshot') ->
profilerBuildSnapshotFile: metadata + raw FCI + exact contents for
every readable file (no ignore list, no masking). The scan returns to
the list afterwards.
- ICCID is decoded from EF.ICCID (2FE2) with decIccid() (nibble-swapped
E.118 digits, F pad per TS 102 221 13.2) and stored as an immutable
snapshot field, shown next to the name in the list and the view.
- Open shows all captured data read-only (attributes, raw FCI + decoded
FCI, contents or 'Not captured'); only the name is editable/saveable.
- Import/Export/Delete + profilerValidateSnapshot; quota-safe save.
Tests: decIccid, profilerSnapshotIccid, validation, snapshot builder
(exact contents/no mask/unreadable), snapshot-mode scan with ICCID.
Docs synced (help/help-ru). SW cache v79 -> v80.
Adds an 'All' header checkbox next to 'Ignore contents of files:' that
checks or unchecks every file; it reflects the list state (checked only
when all are checked, indeterminate on a mixed selection) and updates on
individual checkbox changes. RU label: 'Все'. Docs synced, tests for the
toggle/state helpers. SW cache v77 -> v78.
fcpDecode now parses with a partial-aware walker (fcpParseTlvs) that keeps
every complete TLV it encounters and reports why it stopped:
- 'TLV 62 declares N bytes, only M available' (truncation)
- 'Truncated length field at offset X' / 'incomplete TLV header'
- 'Invalid length form' / 'Trailing data after TLV 62'
- inner A5/C6 errors prefixed with their context ('In A5: ...')
The editor preview shows the decoded parameters plus a red
'Decode failed: <reason>' line (RU: 'Ошибка декодирования'), and the check
report's decoded FCI diff appends per-side error notes (expected/actual)
while still showing whatever decoded on either side.
SW cache v74 -> v75.
parseBerLen treated the first long-form length byte (0x81/0x82/...) as the
count of length bytes, so any TLV with a long-form length parsed as
garbage: the profiler FCI decoder returned ok:false (decoded preview
disappeared and never came back after editing an FCP with a '62 81 xx'
outer length or a long-form inner TLV), and the same bug hit parseTlvList
(C-APDU parser, INSTALL param walker), parseBerScript (expanded script
rows >= 128 bytes) and readLvField.
Per ISO 7816-4 5.2 / UICC_SPECS.md 1.7 the long form is 81-84 followed by
(b & 0x7F) length bytes. Fixed; short-form behavior unchanged.
New tests: parseBerLen short/81/82 forms, parseTlvList long-form outer and
inner TLVs, fcpDecode long-form regression (81/82, nested A5, >=128-byte
FCP) incl. the editor preview path. SW cache v73 -> v74.
The exact-FCI row is now a two-column flex layout: the editable raw FCI
textarea on the left, the decoded FCP list on the right (50/50, textarea
4 rows). Preview rendering split into profilerFciPreviewItems; a new
profilerFciInput hook updates the rule and re-decodes the panel on every
keystroke (no editor re-render, no focus loss), showing an empty body
while the hex is incomplete/invalid. Docs wording updated. SW cache
v72 -> v73.
Adds a spec-verified FCP/FCI decoder (ISO 7816-4 5.3.3 Tables 12-14,
TS 102 221 11.1.1.4; cross-checked against pySim ts_102_221.py — spec wins
on the data coding byte and termination mask). fcpDecode accepts the FCP
template '62', an FCI '6F' wrapper and bare FCP content, and decodes:
file size/total size, file descriptor (access/shareable, file type,
structure incl. BER-TLV/SIMPLE-TLV, data coding byte, record length/count),
FID, DF name, SFI, life cycle status, security attributes, the A5
proprietary sub-DOs (UICC characteristics, power, clock, memory, file
details, sizes, commands, environmental conditions, test config) and the C6
PIN status template DO. Unknown TLVs are preserved raw.
Usage:
- check report: an 'Exact FCI' mismatch now renders a decoded
expected/actual table under the raw fields, highlighting differing
parameters and showing missing ones as '—'
- rule editor: the FCI hex textarea shows a live decoded FCP preview
10 new tests (vectors incl. pySim linear-fixed, malformed inputs, diff
highlighting, report integration). Help docs updated. SW cache v71 -> v72.
Path, FCP/FCI check, file type and size (or record length + count) now
share one flex-wrap row; path column narrowed from w-72 to w-48. The FCI
hex textarea (Exact FCI mode) keeps its own full-width row and the
contents check stays on the following row. SW cache v70 -> v71.
The scan dialog now shows a single 'Ignore contents of files:' label and a
two-column checkbox grid with just the file names (the per-row 'Ignore
contents of ...' prefix was redundant). Drop the now-unused i18n key and
add the new label (RU: 'Игнорировать содержимое файлов:'). Help docs
updated. SW cache v67 -> v68.
The 'matching records: N-M' note was suppressed when every compared record
matched but the record count differed (e.g. 30 expected vs 10 read): the
renderer only looked for failed per-record checks, so the report showed
just the count lines. profilerRunRule now sets res.recordsMismatch whenever
the count or any record differs, and the renderer uses it (with the old
per-record check as fallback), so the matching-record ranges are listed in
that case too. Tests cover the count-only mismatch and the per-record case.
SW cache v65 -> v66.
The Russian help referred to the English i18n keys even though the app
translates those controls. Replaced every UI-control reference with the
label a Russian user actually sees (buttons, pills, tabs, toggles, field
labels): Новый профиль / Профиль с карты / Импорт профиля, Добавить
правило, Сохранить, Проверить, Экспорт, Удалить, Подключиться /
Подключить карту, Отправить на карту, Проверить в pySim, Отправить STATUS,
Опрос, Декодировать, Упаковать в Secured packet, Экспорт/Импорт JSON
buttons, Карты, Файловый менеджер, Пользовательские файлы, Профайлер,
Командная строка pySim, Отправка APDU, Проактивный UICC, Парсер ответов,
Картридер, Разбор C-APDU, справка, Меню STK, Триггер (Push SMS),
Обёртка в Command Scripting template ('AA'), mode labels and scan options.
Also translated the section headings that name UI views (2.6, 3.2, 4, 5,
5.1-5.6) and the proactive-log 'Ответ:' line. Terms without a Russian
translation (SIM RFM, Expanded Script, Secured Packet, RAM, Silent
(P2=0C), event names, spec terms) stay English as in the UI.
Both files: fixed the documented '+ Command' button that does not exist
(the chain builder's buttons are + SELECT, + READ RECORD, ...).
SW cache v63 -> v64.
Audited both help files against the UI and fixed stale/missing content:
Fixed:
- Expanded Script: remove the nonexistent 'Response Type' TLV row; Error
Action is 'one of four forms', not three
- 'Expanded Remote Response' section described a UI that does not exist;
rewritten as response decoding (PoR shown in the Secured Packet view,
last SW/data filled into the Response parser)
- Cards/Custom files: document all five export/import buttons
- Reader auto-detection: drop the stale /dev/ttyUSB0 start.sh bullet, fix
'Reader: none' and the 'Equip card' button name, note the server-side
PC/SC probe with retries
- Profiler: 'Check contents' label, symbolic names next to rule paths,
Add rule/Save buttons, aligned read-only expected/actual fields for
raw-data mismatches
Added:
- 1.1 Interface (header chrome: INSTALL PWA, github/help links, EN/RU and
theme toggles, localStorage persistence, help deep-links)
- GPC v2.2 Amendment B v1.1 in the standards list
- Card reader connect workflow (URL, Connect, status, Equip card)
- File manager tree browser + Save/Cancel
- Send to Card PoR behavior in 3.1/4
Also fixed the app help anchors: C-APDU Parser now opens #c-apdu-parser
and the Profiler sub-tab opens #profiler. SW cache v60 -> v61.
FCI and content (incl. record) mismatches are now rendered as two read-only
monospace inputs — expected on top, actual directly beneath — sharing a
fixed-width right-aligned label column (w-24) so both fields start at
exactly the same horizontal position and stretch to fill the row
(flex-1/min-w-0, no overlap). A title attribute shows the full value on
hover. Non-raw checks (fileSize, recordLen, numRecords, content.records,
read failure) keep the inline line.
New profilerRawDataCheck helper; 2 new tests. SW cache v59 -> v60.
A record file with both a numRecords metadata check and record content
checks reported the same count mismatch twice (numRecords and
content.records). profilerRunRule now omits the content.records line when
the numRecords check already covers it, keeping it only when no numRecords
check ran (FCP/FCI 'type' mode or numRecords cleared) or when numRecords
passes but the read returns a different count.
3 new tests. SW cache v58 -> v59.
Profile-from-card scan form gains two checked-by-default options
'Match first 4 bytes for EF.IMSI / EF.ICCID', surfacing the previously
hardcoded mask behavior. Unchecking captures those files' contents
exactly. profilerBuildFileRule now takes a maskFids set (undefined keeps
the legacy mask default); scan start collects it and threads it through
profilerScanCard.
While a scan runs the options form (ignore list, mask options, FCP/FCI
selector) is hidden, Cancel and Scan are disabled (with visible
disabled styling), and the profile-name input becomes read-only — fixing
the latent cancel-mid-scan bug that left the scan running and popped the
editor open. profilerFromCard performs a full fresh-open reset.
4 new tests (mask/exact/legacy + scan threading). RU i18n key, help docs
synced. SW cache v57 -> v58.
The Check report now states exactly what was verified for each file:
- passing files append a plain summary to the path line
('filetype and size, contents' / 'filetype and records, contents' /
'Exact FCI, contents')
- mixed results mark each aspect inline
('filetype ✓, size ✗, contents ✓'), with the existing red detail lines
for the mismatches; 'Exact FCI' subsumes type/size/records
- record files with a contents mismatch add a 'matching records: 1-5,
7-10' note listing the records that did match
New pure helpers: profilerResultAspects (derives checked aspects from the
result checks), profilerAspectSummary (plain vs marked text),
profilerNumRanges (compresses record numbers into ranges). profilerRunRule
now also returns res.recordsMatched. 6 new tests incl. an end-to-end report
render. RU i18n keys added. Help docs synced. SW cache v56 -> v57.
- Path field narrowed to w-72 with the FCP/FCI selector to its right in
one row
- File type + size (or record length + count) now sit in a single flex
row (fixes the accidental wrap from the unavailable md:grid-cols-4 class
in the prebuilt CSS)
- 'Contents' label renamed to 'Check contents' and the None/Exact/Mask
radios moved inline to its right, saving one line per rule
Rendering-only change; no logic or test behavior affected. SW cache v54 -> v55.
Split profilerScanCard into two phases so the file total is known up
front: phase 1 walks the filesystem (tree calls only) collecting every
file entry, phase 2 builds a rule per file. The optional onProgress(done,
total, path) callback reports 'N / total files (%) — path'; the scan
dialog shows it under the FCP/FCI selector, starting with 'Discovering…'
during phase 1. No extra card I/O (tree calls are no longer interleaved
with select/read). Backward compatible: onProgress is optional.
2 new tests (progress sequence 0/total then 1..total with paths; scan
without callback). RU i18n keys (Discovering..., files). Help docs synced.
SW cache v53 -> v54.
Each rule now carries an fciMode ('type' | 'type_size' | 'exact') plus an
fciHex (raw SELECT response) so rules can verify different depths of file
control information:
- Filetype only (FCP): exists + file type
- Filetype + size (FCP): adds file size (or record length/count) - previous
behavior
- Exact FCI: adds byte-for-byte comparison of the raw SELECT response (the
FCP '62' template), catching FID/AID, life-cycle, security-attribute and
proprietary-parameter changes
/profile from card/ gains a matching FCP/FCI selector (default Filetype +
size); rules store fciMode and always capture fciHex so they can be
upgraded to Exact FCI in the editor without rescanning. The rule editor
adds the selector, hides size/record fields in 'type' mode and shows an
editable FCI hex textarea in 'exact' mode. Contents checks stay independent.
Server /api/select now returns fci_hex (raw FCP template hex, uppercased).
profilerValidateProfile accepts the new mode; profilerNormHexStrict added
for byte-exact comparison (no '?' wildcards). Legacy rules without fciMode
default to type_size.
10 new tests (build-rule fields, run-rule modes incl. byte compare +
missing-FCI, validation). Docs + RU i18n synced. SW cache v52 -> v53,
version 1.9.28.
Rules now persist the pySim symbolic name captured during 'Profile from
card' (profilerBuildFileRule stores c.name || sel.name), and the profile
edit view renders it next to the Path label using the same lookup order
as the check view: profilerCustomNameForPath(path) || rule.name.
Path edits call the new profilerUpdateRulePath, which drops the stale
scan-time name and refreshes the label from the custom-files dictionary
live, so the displayed name always corresponds to the current path.
Manually added rules get name: null (resolved via custom-file lookup).
Old profiles without rule.name still work.
4 new tests (rule.name from child/select/null, custom name lookup, path
edit clears name + refreshes label). SW cache v51 -> v52.
The 'Profile from card' ignore checkboxes matched only by FID, but the
list had EF.KcGPRS at 4F52 (TS 31.102 DF.GSM-ACCESS) while the live card
exposes it at 6F52 (TS 51.011 DF.GSM, verified in TS 51.011 v4.15.0
10.3.32 and the DF.GSM allocation table). The miss made profilerBuildFileRule
capture full contents (Exact) for a file the user had checked to ignore.
- Correct EF.KcGPRS FID to 6F52 (spec-verified)
- Match ignores by FID OR by pySim name: each checkbox now carries
data-ignore-name, profilerScanStart builds an ignoreNames set, and
profilerBuildFileRule checks both. Covers FID variants (6F52 vs 4F52)
and future constant typos; ignoreNames is optional for back-compat.
- 6 new tests: ignore-list FID/name sanity (incl. KcGPRS=6F52), duplicates,
ignored-by-FID, ignored-by-name-only regression case, non-ignored control,
back-compat when ignoreNames is omitted.
SW cache v50 -> v51.
catarrh