[Kshot3000]

Problem

SigmaByteReader silently invented data when a parse ran past the end of the input:

• readByte() returned undefined (typed as number) at EOF, so readBool() quietly became false and the type serializer's byte.toString(16) crashed with an unrelated TypeError.
• readBytes(n) clamped to whatever remained and advanced the cursor past the end, so fixed-length fields (box/transaction ids, group elements, AVL digests, ErgoTree bodies) parsed as short values instead of failing.
• readBits(n) read missing bytes as 0, inventing false bits for truncated SColl[Boolean] constants.
• A VLQ whose last byte had the continuation bit set decoded to a partial value: readUInt/readBigUInt on [0x80] returned 0. The same applied to any length-prefixed field — readI256 with a declared length of 5 and only 2 bytes remaining returned a wrong bigint instead of an error.

The reference implementation this mirrors (scorex-util VLQByteBufferReader) fails closed here: getByte() is ByteBuffer.get (throws on underflow) and getBytes(size) requires size <= remaining.

Fix

readByte, readBytes, and readBits now throw a RangeError naming the position and input length when the read would go out of bounds (including a negative readBytes length, which previously moved the cursor backwards). This makes truncated VLQs and truncated length-prefixed fields fail at the point of the bad read.

Deliberate non-changes:

• peek()/match() keep their clamping semantics — they don't advance the cursor and the existing tests document peek returning "all bytes if length is greater than available bytes".
• readVLQ/readBigVLQ still return 0 for a reader that is empty before the read starts (their existing explicit isEmpty check); only a VLQ truncated mid-sequence now throws, via readByte.

Tests

Five new tests in sigmaByteReader.spec.ts, all of which fail on pristine master and pass with this change: byte past EOF, byte-slice past the remainder (with the exact remainder still readable), bits past EOF, continuation-truncated VLQs for both readUInt and readBigUInt, and a truncated length-prefixed readI256.

Full monorepo suite: 1137/1137 tests pass (66 files), so no in-repo parse path was relying on the lenient behavior.

---

@Kshot9000 · Donations (ERG): 9fcM5RWnAjmP4vx5bnW6yohB6H9bLq8sJbaPLHtwZLtQPB32Pvy

[changeset-bot]

🦋 Changeset detected

Latest commit: 4f8832f

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@fleet-sdk/serializer	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR