simple/CHANGELOG.md

# Changelog

All notable changes to SIMple (the PWA and `pysim-simple-server`).
The format follows [Keep a Changelog](https://keepachangelog.com/);
versions use `MAJOR.MINOR.PATCH` (see the README).

History before the first packaged release (v3.31.1) lives in the git log —
every change carries a descriptive commit message.

## [Unreleased]

## [3.35.0] — 2026-10-23

### Added

- **EF.SMSP decoded editor** — the full SMS-parameter form (TS 31.102 §4.2.27): alpha,
  TP-Destination Address and TS-Service Centre Address (each with its own length
  convention and a TON select; an empty SCA means the default SMSC), TP-PID, TP-DCS and
  the relative TP-VP — every parameter with an *absent (the ME default applies)* toggle.
  Live hints decode the DCS (TS 23.038 §4: coding group, alphabet, class, compression,
  message-waiting) and the relative validity period (TS 23.040 §9.2.3.12.1).  Unchanged
  parameters keep their exact bytes (a card quirk such as a DCS marked absent with byte
  `00` survives), alphanumeric TON addresses are preserved read-only, and all checked
  records are validated before anything is written.
- **Edit decoded, wave 2** — new editors for the service tables **EF.UST / EF.IST /
  EF.EST** (a filterable bit grid preserving services the table does not model), the
  scalars **EF.ACC**, **EF.PHASE**, **EF.HPPLMN** and **EF.AD**, the PLMN lists
  **EF.PLMNsel / EF.FPLMN / EF.EHPLMN**, the identities **EF.ICCID / EF.IMSI** (behind a
  confirmation — writing them changes the card identity) and the cell-broadcast lists
  **EF.CBMI / EF.CBMID / EF.CBMIR**.  **EF.SST** is deliberately left read-only: its
  SIM layout is 2 bits per service (allocated + activated), which needs its own editor.
- The file-format survey (`docs/decoded-editing.md`) now records the shipped waves and
  marks **EF.DIR / EF.ARR / EF.SMS / EF.PBR** as postponed by decision (read-only).

### Fixed

- `efDecSmsp` mapped the parameter indicators shifted (SCA on bit 5, PID on bit 2, DCS
  on bit 3, VP on bit 4); the spec order is DA/SCA/PID/DCS/VP on bits 1–5 with bits 6–8
  reserved — a record with a mixed indicator (e.g. `E9`) showed DCS and hid VP.
- The read-only EF.SMSP view decoded the TP-DA length as octets, although it is the digit
  count (TS 23.040 §9.1.2.5); a slot padded with `00` gained phantom digits.
- The **Read decoded** pill no longer looks disabled when it is enabled: an inactive-but-
  enabled pill is now an outlined button (blue on white/slate), and the grey fill and
  dimming are left to the actually disabled state.
- **EF.ACC** class numbering (TS 31.102 4.2.15): byte 1 b8…b1 codes classes 15…8 and
  byte 2 classes 7…0 — the decoder counted class 0 at the MSB (an all-FF ACC read as
  class 0 instead of classes 0–15).
- **EF.AD** (TS 31.102 4.2.18): the operation mode is a whole-byte enum
  (00/80/01/81/02/04), not bits 0-1 plus a test bit; the MNC length may be 0 (service
  n°130 available) or absent (3-byte SIM form) — the editor previously refused both.
- **EF.SST** (TS 51.011 10.3.7): the SIM service table is decoded with its 2-bits-per-
  service layout (allocated + activated) instead of the UST 1-bit codec.
- **EF.IMSI**: the parity/type nibble is `9` for an odd digit count and `1` for an even
  one (TS 24.008 10.5.1.4); the encoder always wrote `9`.
- Saving a PLMN/CB list with no edits no longer rewrites it as all-FF (the collected rows
  and the list encoders disagreed on the value shape).

## [3.33.0] — 2026-10-23

### Added

- **Edit decoded** in the file manager: a file is read raw, decoded into its own editable
  form, re-encoded in the browser and written through the normal update path.  The first
  wave covers **EF.SPN**, **EF.PLMNwAcT / EF.OPLMNwAcT / EF.HPLMNwAcT**, the dialling-number
  files **EF.ADN / EF.FDN / EF.SDN / EF.MSISDN / EF.LND / EF.BDN** and **EF.PNN**.  Shapes:
  a flat transparent form (SPN), a transparent list editor with add/remove rows (PLMN+ACT),
  per-record forms with a selection checkbox that is auto-checked when a field changes
  (dialling numbers, PNN) and per-record TLV item rows with add/remove/edit (PNN's full and
  short names plus the PLMN additional information).  All checked records are encoded and
  validated before anything is written; unchanged fields, record lengths, extension
  pointers (EF.EXT1–EXT5), the BDN comparison-method byte and unknown PNN TLVs are
  preserved, and an edit that would break a reference (a number continued in an extension
  file, a dropped mandatory PNN full name) is refused.
- `docs/decoded-editing.md`: the file-format survey behind the design (every known EF
  classified by structure/coding into the editor's shapes S0–S10, with the release waves).

### Removed

- The file manager's decoded view no longer carries the collapsed **pySim JSON (server)**
  disclosure — the view is the client-side EF decoder's table only.  The disclosure was an
  unrelated *second* decode: opening it issued another `/api/read` with `mode: 'decoded'`
  to run pySim's `read_binary_decoded`/`read_records_decoded`, and it was reachable only
  when our decoder covered the file (`Read decoded` is decoder-gated), so the one case
  where a second opinion would matter — a file our decoders do not cover — never offered
  it.  `/api/read` keeps `mode: 'decoded'` for API callers; the help, READMEs and project
  notes say the decoded table is built in the browser from the same raw bytes.

### Fixed

- **EF.BDN** decodes and encodes with its own `X+15` layout (the trailing Comparison Method
  Pointer); the shared ADN codec had treated it as `X+14`.
- `efPlmnAct` decoded the EC-GSM-IoT-only ACT (`b8=1,b4=1,b3=0`, `0x0088`) as nothing while
  checking the impossible `0x0086` (TS 31.102 Table 4.2.5.2).
- PNN network names: the decoder now trusts a non-zero spare-bit count before the
  "all bytes < 0x80 → uncompressed" heuristic, so a valid packed name can no longer be
  misread (and later written back wrong).
- `/api/write` and `/api/read` report a cmd2 exception in the captured output as a failure
  instead of `success: true` with `SW: 9000`.
- The empty amber **preset-mismatch banner** no longer flashes on the SCP80/RAM view.
  `ramOpChanged()` (v3.28.0) un-hid `#ram-preset-warning` by the operation kind, but the
  banner's text *and* visibility belong to `spPresetWarningRender()` (which also handles
  the batch suppression): with no warning text the empty bordered box stayed visible until
  the next status poll re-hid it — on every RAM view and operation switch.  The visibility
  is content-driven again; `frontend/tests/ram.test.js` pins it (the pre-fix code fails
  the test).

## [3.32.5] — 2026-10-21

### Fixed

- A card removed during the **startup initialization** no longer leaves the server
  reporting a phantom equipped card.  The init caught the PC/SC error (and printed a
  traceback) but kept the card object it had already built, so `card_present` and
  `_CARD_CONNECTED` stayed True — and the auto-equip then skipped every later insertion
  (its `_CARD_CONNECTED` early return), leaving only a manual Equip or a restart.  A
  PC/SC error now prints one line (`INIT: card removed during initialization …`), drops
  the half-initialized session and lets the presence monitor equip the next card.

## [3.32.4] — 2026-10-21

### Fixed

- The auto-equip's captured traceback now always includes the **exception line**: cmd2
  renders a rich-box traceback longer than the logged head, so the actual error was cut
  off.  The tail is appended after a `...` marker; the "releasing the old link failed"
  message now says it continues with a fresh link (a removed card's dead PC/SC handle
  always fails to disconnect).
- A card insertion now marks the PC/SC link for a rebuild, so the **first** auto-equip
  attempt already uses a fresh link.  On Windows the old link cannot connect after a
  swap — v3.32.2 recovered it, but only on the second attempt, after a failed init, a
  2–4 s delay and a traceback in the log.  A link created at a cardless startup has
  never connected either, which the same mark covers.  The rebuild still runs in the
  equip worker (never in the pyscard monitor thread) and `_ensure_transport` releases
  the old link first.

## [3.32.2] — 2026-10-21

### Fixed

- **The auto-equip loop after a card swap on Windows.** A failed initialization with
  the card still present was reported as "card gone during initialization" and treated
  as a no-op success — so the three retries never ran and the watchdog re-armed every
  few seconds forever.  It is a failure now, and **every failed attempt marks the
  PC/SC transport for a rebuild**, so the retry starts from a fresh link (the release
  of the old link happens first, so it cannot inherit a stale card handle).  The error
  pySim swallowed is printed as `AUTO-EQUIP: init output:` with its traceback.

## [3.32.1] — 2026-10-21

### Added

- **"Start counter value"** in the outfile import: the counter every created keyset
  starts from.  It defaults to the store's value (`1`); raise it for cards whose
  counters are already past it.  The API takes `start_counter` (1–10 hex digits;
  a bad value is a 400), and the preview carries the value.

## [3.32.0] — 2026-10-21

### Added

- **Updating a packaged install** — `update.sh` / `update.bat` (and `tools/update.py`
  beneath them): downloads the latest release package, replaces the application files
  in place and re-installs the app into the **existing `.venv`**.  pysim and its
  prerequisites (the slow part: pyscard builds from source on Linux) are kept, so an
  update takes seconds.  `--check` reports the current and latest versions, `--version`
  picks a specific release, `--zip` updates from a local package (offline), `--no-pip`
  refreshes the files only, `--force` re-applies the same version.
- The release workflow now takes the release notes from this file's matching version
  section (falling back to generated notes when it is missing).

### Changed

- Installs made from the v3.31.x packages have no updater yet: copy `update.sh`,
  `update.bat` and `tools/update.py` from this package once (or re-run `setup` from it).
  Later updates use `update.sh` / `update.bat` — or `git pull` for a clone.

## [3.31.2] — 2026-10-20

### Changed

- The **SCP81 TLS-PSK listener states its Python requirement**: it needs Python 3.13+
  (`ssl`'s PSK callbacks exist only since 3.13).  Below that it raises a clear error
  instead of `AttributeError`, and the PSK tests are skipped (they run on 3.13+).
- The READMEs and the setup scripts: **Windows strongly recommends Python 3.13**
  (pyscard has wheels up to 3.13; on 3.14 pyscard and some pysim dependency wheels are
  unavailable and build from source with MSVC).

## [3.31.1] — 2026-10-20

The first packaged release (`simple-3.31.1.zip`).

### Added

- **EF_PBR (phonebook reference) support** — the PBR is decoded in the decoded file
  views (the `A8`/`A9`/`AA` containers, the `C0`–`CC` kinds, the master EF), and the
  phonebook files it references (they are not in pySim's model) are added to the file
  tree server-side, selectable and readable.
- **In-app dialogs** — the native alert/confirm/prompt dialogs are replaced by a shared
  modal (queued, keyboard-aware, focus-trapped, ARIA-labelled).
- **Windows console encoding guard** — the `start.bat` PowerShell pipe can no longer
  kill the server on a character the ANSI code page lacks.
- **GPL-2.0-or-later licensing** (`LICENSE`, `THIRD-PARTY-NOTICES`), matching pysim.
- The curated **release package** (`tools/build_release.py`) and the tag-triggered
  `.github/workflows/release.yml`.

### Changed

- `setup.sh` / `setup.bat` and the READMEs: on Linux `pyscard` builds from source and
  needs the PC/SC headers (`libpcsclite-dev`).