Package dev.relism.flash.http2.hpack

Class HpackIntegers

java.lang.Object

dev.relism.flash.http2.hpack.HpackIntegers

public final class HpackIntegers
extends Object

RFC 7541 §5.1 prefix-coded integer encode/decode. An N-bit prefix holds values 0..2^N-2 directly; the sentinel 2^N-1 means "the real value is at least this large, keep reading continuation octets" — each contributes 7 bits, low-to-high, with the high bit as a continue flag.

Overflow safety (HPACK bomb)

RFC 7541 places no upper bound on the number of continuation octets — a hostile peer can encode an arbitrarily large integer (conceptually up to 2^64 and beyond) in a handful of bytes. decode(byte[], int, int, int) rejects any integer needing more than MAX_CONTINUATION_OCTETS continuation octets, and independently rejects one that would exceed Integer.MAX_VALUE even within that octet budget — belt-and-suspenders, since with the octet cap in place the second check is not expected to ever fire on a real input. Both throw the preallocated Http2Exception.COMPRESSION_ERROR singleton — zero allocation on this hot rejection path (see that exception's own Javadoc for why reusing a singleton here is safe).

Method Summary

Modifier and Type	Method	Description
static long	decode(byte[] buf, int pos, int limit, int prefixBits)	Decodes a prefix-coded integer starting at buf[pos], where the low prefixBits bits of buf[pos] carry the prefix (any higher bits — e.g. a representation's leading flag bits — are the caller's concern and are masked off here).
static void	encode(ByteWriter out, int prefixByteFlags, int prefixBits, int value)	Encodes value as a prefix-coded integer into prefixByteFlags | encoded value, writing into out.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

decode

public static long decode(byte[] buf,
 int pos,
 int limit,
 int prefixBits)

Decodes a prefix-coded integer starting at buf[pos], where the low prefixBits bits of buf[pos] carry the prefix (any higher bits — e.g. a representation's leading flag bits — are the caller's concern and are masked off here). limit is the exclusive end of the region this integer may read from (typically the end of the current HPACK block) — reading past it means a truncated/malformed encoding, not "need more input", since by the time this runs the whole block is already contiguous in memory (RFC 9113 §6.10: CONTINUATION frames are never interleaved with other frames).

Returns:

Pairs.pack(int, int)(value, newPos) — the decoded value in the high 32 bits, the position just past the last consumed byte in the low 32 bits

Throws:

Http2Exception - COMPRESSION_ERROR on truncation, on exceeding MAX_CONTINUATION_OCTETS, or on a value that would exceed Integer.MAX_VALUE

encode

public static void encode(ByteWriter out,
 int prefixByteFlags,
 int prefixBits,
 int value)

Encodes value as a prefix-coded integer into prefixByteFlags | encoded value, writing into out. prefixByteFlags carries whatever high bits the representation needs (e.g. 0x80 for an Indexed Header Field) already shifted into position — this method only ever sets the low prefixBits bits of the first byte.