[tannevaled]

Closes #787.

What it was

A YAML-loaded object answered o.a == 1 and instance_variables == [].

That is the worst version of this bug. The ivars were there and readable by
name, so every direct check passes; what breaks is anything that walks the
object — marshalling, inspect, a serialiser, an equality written over
instance_variables — and it breaks by seeing an empty object, silently.

Cause

For an *RObject, ivarStoreOf returns order: &o.ivarOrder — the address
of a struct field, so the order pointer is never nil and ivarNamesInOrder
always takes the ordered branch. A constructor that fills the ivars map directly
and leaves ivarOrder empty therefore produces ivars that read back and do not
enumerate, with nothing anywhere to say so.

exception.go:377 builds its RObject with an ivarOrder, which is why
exceptions enumerate correctly and this stayed invisible.

Two changes, because one is the fix and the other is a floor

yaml_bind.go convObject now fills ivarOrder alongside ivars, in the
sequence the yaml package documents for emission (value.go:86-89 — the names
in o.Order that exist, then the rest lexicographically). That is the
document's order, so the result matches MRI:

document z, a, m	MRI [:@z, :@a, :@m]
rbgo after	[:@z, :@a, :@m], identical over three runs

z, a, m is neither alphabetical nor reversed, and Go's map iteration is
randomised per run, so reproducing it by accident is not plausible.

ivarNamesInOrder now also reports live names missing from the order list,
sorted. Constructors that fill ivarOrder never reach that path, so creation
order is preserved where it is known; what it buys is that the next
constructor to fill a map directly is wrong visibly rather than silently.

What the coverage gate asked, and the two different answers

The gate named two functions, and they needed opposite treatment — which is
exactly why it is worth it asking rather than me deciding in advance.

• ivarNamesInOrder's new arm is unreached, not dead. Nothing gets there any
  more precisely because the constructor was fixed. It guards a shape any new
  code can produce in one line, so it gets an internal test that builds that
  shape: an empty order list (all names, sorted), a half-filled one (the listed
  name keeps its place, the rest follow sorted), and a name in the list whose
  value is gone — which stays gone, since a floor that resurrected a removed
  ivar would be worse than no floor.
• convObject's missing-key arm is dead: yamlIVarOrder returns only keys
  present in o.IVars, so the lookup cannot fail. Removed rather than tested,
  because a test for an unreachable branch asserts nothing.

A note on the writing of it

yamlIVarOrder already existed in the same file, eleven lines from where I
wrote a second copy of it. The duplicate is not in this branch.

🤖 Generated with Claude Code