zeekay d871397e3d The Q-chain reads its wire out of the schema too ...

Three structs — the signature preimage, the envelope over it, and the
block — stated their offsets by hand in wire.hpp and used them by hand in
wire.cpp. They are now schema/wire.zap, and zapgen prints the readers and
the builders. 340 lines become 43 of schema and what is left of
wire.cpp, which is the canonicality rule, the size bound, and the walk
over the length vector — the things that are about a Q-chain block rather
than about a buffer.

One behaviour changed, and it is the one the schema buys. TxLens is a run
of u32, so the generated reader takes the stride and the wire layer now
refuses a length word the message cannot hold — where the strideless read
accepted it and left every element accessor answering zero. The chain's
own bound still names a count the BLOB cannot back; the wire's bound now
names one the MESSAGE cannot hold, and parse_tx_set tells the two apart
by the absent list rather than losing the reason to a later
non-canonical. Both refuse; both say why.

The regeneration target moved into zap.cmake as zap_schema(), so a chain
declares its schema in one line and no chain spells the generator
invocation twice.

Green, and the two that matter are Go's own bytes:
AnUnsignedTransactionSerializesAsGoDoes and
ABlockCarryingAnUnsignedTransactionHasGosID. 13/13, and xvm still 59/59.

2026-09-06 12:44:56 -07:00

..

wire.zap

The Q-chain reads its wire out of the schema too

2026-09-06 12:44:56 -07:00