forgertele 1.0.0__tar.gz

forgertele-1.0.0/ACKNOWLEDGMENTS.md

@@ -0,0 +1,108 @@
+# Acknowledgments
+
+`opentele-ng` is a fork of [thedemons/opentele](https://github.com/thedemons/opentele),
+released under MIT. The 11-phase modernization (May 2026) integrated work
+from many sources — credit where credit is due.
+
+## Upstream
+
+- **[thedemons](https://github.com/thedemons)** — original `opentele`
+  library (2022). All core architecture (`TDesktop`, `Account`, `MapData`,
+  `Storage`, `MTProto auth`) is from upstream. Without it, this fork would
+  not exist.
+
+## Contributing forks (integrated patches)
+
+The fork analysis surveyed 40+ active forks of upstream; the following
+contributed actual patches or specific discoveries that landed in
+`opentele-ng`:
+
+### Phase 1 — Python 3.13/3.14 + PyQt6
+- **[azamtoiri/opentele](https://github.com/azamtoiri/opentele)** — Python 3.13
+  fix approach: `crossDelete` extension + `__func__`-based bound method
+  comparison.
+- **[iamlostshe/opentele](https://github.com/iamlostshe/opentele)** (deep fork
+  via AlreadyNobody) — pointed to `tgcrypto-pyrofork` as replacement for
+  abandoned `tgcrypto`.
+- **[hustLer2k/opentele](https://github.com/hustLer2k/opentele)** (deep fork)
+  — UTF-8 encoding for `setup.py` README read (Windows fix).
+- **[Snowing](https://github.com/Snowing)** — `account.py:186` fix: invalid
+  `map.stream >> legacyBackgroundKeyDay` (Python no-op) replaced with
+  `readUInt64()`.
+
+### Phase 1.5 — Wire-format fixes (verified against TDesktop C++)
+- **[RobertAzovski/opentele](https://github.com/RobertAzovski/opentele)** —
+  discovered 4 new `lskType` keys for Telegram Desktop 5.x-6.x tdata format:
+  - `lskRoundPlaceholder` (0x1A) — single uint64
+  - `lskInlineBotsDownloads` (0x1B) — single uint64
+  - `lskMediaLastPlaybackPositions` (0x1C) — single uint64
+  - `lskBotStorages` (0x1D) — corrected wire format to
+    `Dict[PeerId, FileKey]` map after 3-AI code review (initial submission
+    was a single uint64).
+- `lskPrefs` (0x1E) — found by Codex during AI review by direct source dive
+  into upstream Telegram Desktop C++ — not present in any community fork.
+
+### Phase 2 — 2026 device fingerprints
+- **[Paramon/opentele](https://github.com/Paramon/opentele)** — base for
+  modern macOS device list + `_generate_tdesktop_app_version(unique_id)`.
+- 2026 updates (Galaxy S26 series, M5 Mac lineup, iPhone 17 / Air, Android
+  SDK 36/37) added during the Phase 2 work itself with verified sources.
+
+### Phase 3 — Multi-account + QoL
+- **[snakechilds/opentele-nuitka](https://github.com/snakechilds/opentele-nuitka)**
+  — `kMaxAccounts` bump, `**kwargs` forwarding pattern in `FromTelethon`,
+  Nuitka-friendly `sharemethod` rewrite.
+- **[Ehekatech/opentele-tg](https://github.com/Ehekatech/opentele-tg)** —
+  QR-login documentation skeleton.
+
+### Phase 6 — Pyrogram bridge + layer-aware fingerprints
+- **[timka-123/opentele](https://github.com/timka-123/opentele)** — the idea of
+  a Pyrogram bridge (`FromPyrogram` / `ToPyrogram`), which upstream listed as an
+  incoming feature and never shipped. The implementation here is written from
+  scratch: theirs does not run (`account` used before assignment, `writeKeys`
+  stamping `MainDcId` onto every key, `session_name=` which Pyrogram 2.x does
+  not accept), which their README acknowledges.
+- **[anmv/opentele](https://github.com/anmv/opentele)** — the insight that the
+  advertised Telegram Desktop version should match the MTProto layer the
+  client actually speaks. They hardcode the version/layer pair; `opentele-ng`
+  derives it from the installed Telethon at runtime
+  (`TELEGRAM_DESKTOP_LAYERS`).
+
+## Telegram Desktop source
+
+Wire format for the new `lskType` keys, `_settingsKey` AES alignment fix
+(1.0.1), and DoS guard upper bounds (1.0.3) verified against upstream
+**[telegramdesktop/tdesktop](https://github.com/telegramdesktop/tdesktop)**
+`Telegram/SourceFiles/storage/storage_account.cpp` and
+`storage_file_utilities.cpp` (dev branch).
+
+## Code review
+
+Every phase of `opentele-ng` was reviewed by **three independent AI systems**
+running in parallel before push: **OpenAI Codex**, **Cursor**, and **Google
+Gemini**. They caught critical issues that would otherwise have shipped:
+
+- Phase 1.5: `lskWebviewTokens` was uint64×2 (wrong; should be QByteArray×2)
+  — caught by all three AI; `lskBotStorages` was a single uint64 (wrong;
+  should be a map) — Cursor + Codex; `lskPrefs` was completely missing —
+  Codex via C++ source dive.
+- Phase 2.5: `AndroidDevice.__gen__` was iterating the legacy 4500-model
+  list against modern Android 13-17 SDKs (Galaxy S4 + Android 17 fingerprint)
+  — all three; `macOSDevice.__gen__` `FromIdentifier` parser was corrupting
+  clean model names (`"Mac16,9 (MacBook Air M4 13-inch)"` →
+  `"MacMacBook Air Minch"`) — all three; macOS versions lost the `"macOS "`
+  prefix — Codex.
+- Phase 4: pure-Python `qdatastream` had no direct test coverage (tests used
+  PyQt6 as oracle, never exercising the new code) — Codex; QString goldens
+  missing — Gemini.
+- Phase 1.0.2: untrusted `count` fields in `MapData.read` count loops were
+  a real DoS vector — Codex's prediction confirmed by hand-built malformed
+  payloads; this became the 1.0.3 release.
+- Phase 1.0.3: a third count-loop in `tdesktop.py:__loadFromTData` was
+  missed by the first pass — Codex + Gemini; the in-test `pytest.xfail()`
+  fallback was a fail-open canary that needed to become a hard assert
+  after the fix — all three.
+
+## License
+
+MIT (same as upstream). See [LICENSE](LICENSE).