Fix TypeScript/ESLint errors, add unit tests, and fix failing SRP tests

[Moustafa-Elgammal]

Summary

• Resolve all TypeScript strict-mode and ESLint errors introduced by the TypeScript 5.7 upgrade (stricter Uint8Array<ArrayBufferLike> generics and await-thenable lint rule)
• Add a Docker-based test runner and structured unit tests covering each fix category
• Fix two long-failing SRP (2FA) tests by correcting stale mock test vectors and making the crypto subtle object robust against test-environment timing

---

TypeScript & ESLint fixes (47 files)

Uint8Array<ArrayBufferLike> → explicit generic casts

TypeScript 5.7 narrows Uint8Array to carry its buffer type as a generic parameter. Many call sites passed Uint8Array<ArrayBufferLike> where Uint8Array<ArrayBuffer> (or BlobPart[], BufferSource, etc.) was expected. Fixed with targeted as casts at each boundary:

• blobConstruct.ts — cast blobParts to BlobPart[] for the Blob constructor
• tl_utils.ts — cast buffer.buffer to ArrayBuffer in TLDeserialization, fix fetchRawBytes overload signatures, and cast gzipUncompress result to Uint8Array<ArrayBuffer>
• srp.ts — cast buffer arguments to Uint8Array<ArrayBuffer> for invokeCrypto('pbkdf2', ...)
• passcode/utils.ts — cast salt to Uint8Array<ArrayBuffer> for WebCrypto PBKDF2 params
• serviceWorker/stream.ts, rtmp.ts — cast body bytes to Uint8Array<ArrayBuffer> for Response constructor
• sha1.ts, sha256.ts, aesCtrUtils.ts, aesCTR.ts, aesLocal.ts — cast digest/encrypt inputs
• p2PEncryptor.ts — cast encrypted buffer slices
• 30+ component files — cast Promise.all([...MaybePromise]) arrays with as Promise<T>[] to satisfy the await-thenable ESLint rule

Array.prototype.at(-1) → index replacement

starGiftInfo.tsx used .at(-1) which is not available in the es2015 compilation target. Replaced with arr[arr.length - 1].

---

subtle.ts — lazy crypto proxy

The original module captured window.crypto.subtle at import time:

// before
const subtle = typeof(window) !== 'undefined' && 'crypto' in window
  ? window.crypto.subtle : self.crypto.subtle;

In the vitest/jsdom environment (with isolate: false), modules are evaluated during test collection — before setupFiles installs self.crypto. This could cause crypto operations to bind to an uninitialised or wrong subtle object.

Replaced with a Proxy that resolves getSubtle() on every call:

// after
const getSubtle = () => window.crypto?.subtle ?? self.crypto.subtle;
const subtle = new Proxy({} as SubtleCrypto, {
  get(_t, prop) { return (...args) => getSubtle()[prop](...args); }
});

---

SRP mock test vectors corrected

src/tests/srp.test.ts had two failing tests since the passwordHashed and M1 values in src/mock/srp.ts were stale. A standalone Node.js webcrypto script independently computed the correct values by following the Telegram SRP spec step-by-step (SHA-256 → SHA-256 → PBKDF2-SHA512 × 100 000 → SHA-256). The code was verified correct; only the mock was wrong.

Updated:

• passwordHashed — correct 32-byte SHA-256 output of the full password-hashing chain
• M1 — correct 32-byte proof-of-password value derived from the updated hash

A matched the existing mock exactly, confirming BigInteger modular exponentiation is correct.

---

Test infrastructure

Docker test runner (docker-compose.yaml) — added tweb.test service that runs pnpm test inside the container, enabling reproducible CI-style test execution.

Unit tests — replaced the monolithic fixes.test.ts with four focused files:

File	Coverage
blobConstruct.test.ts	blobConstruct with various BlobPart inputs
bufferConcats.test.ts	bufferConcats + Uint8Array.prototype.concat polyfill
promiseAll.test.ts	Promise.all with mixed MaybePromise arrays
typeCasting.test.ts	Array last-element indexing, ArrayBuffer casts, Response body casts

crypto_methods.test.ts — the existing pbkdf2 test had an empty body (all assertions were commented out). Replaced with assertions that verify output length (64 bytes / SHA-512) and determinism.

---

Test results

Test Files  11 passed (11)
     Tests  70 passed (70)

[morethanwords]

Thank you for the PR! I'm actually using typescript@5.8.2 (as in pnpm-lock.yaml), so the issues that you fixed here is not actual, both Promise and Uint8Array. However, I could merge the tests part of the PR, if you create it again.