[bfrancojr]

Fixes #141.

sbe::Codec::New now refuses a field in a oneof, in a message or a composite. It uses protowire-go's texts (encoding/sbe/template.go, present since d137dd0):

• in a message: sbe: oneof field <M>.<f> not supported;
• in a composite: composite <C> contains oneof field <f>, under the caller's prefix, for example sbe: composite M.c: composite C contains oneof field a.

What still binds: a proto3 optional field, which sits in a synthetic oneof, and a oneof in a group entry. Go does not check either.

This narrows what binds. Before, such a message bound with a block slot for every member, and Marshal wrote the unset members as zero. The CHANGELOG says so.

Merge order: #123 → #124 → #137 → #140 → #143 → this PR. The base is sbe-encoding-range-133 (#143), because the check sits in the same bind code.

Evidence (macOS, fresh build at d00be54):

• Measured before the fix. With the new test SbeOneof.NewCodec and no check, all four shapes Go refuses bound in C++ (bound; want …). These were a message, a composite, a nested composite, and a composite in a group. The three shapes that must bind did bind. This run is also the mutation check: without the check, each refused case fails.
• After the fix: SbeOneof.NewCodec passes, and ctest passes 690 of 690.
• Differential. scripts/go_differential.sh against protowire-go main 207fdfc and spec main 0317811 exits 1. Every section ran: 954 damaged streams decoded alike, 18 SBE schemas converted alike, 9 of them refused, 362 documents. The 450 differing lines are PXF only, the same count as on sbe: refuse a value its (sbe.encoding) primitive cannot hold (#133) #143: pxf: reject null on a (pxf.required) or @required field #131's null rule, pxf: FormatDocument drops every directive, so a @dataset document formats to nothing #130's formatter and pxf: Marshal leaves out an unset (pxf.default) or (pxf.required) field, so it doesn't read back #129. Those have their own PRs (pxf: reject null on a (pxf.required) or @required field #138, pxf: keep every directive when formatting a document (#130) #135, pxf: write an unset annotated field so that it reads back (#129) #136). No SBE line differs.
• Lint: clang-format 18.1.8 is applied. clang-tidy shows no hit in template.cc.

Go has no test for this check, so the test cases come from the issue and Go's code.

[bfrancojr]

Rebased onto main e1c1b72 (#136 merged); head is now e45944c. The PR's own diff is unchanged, checked by comparing its added and removed lines before and after the rebase. Rebased as part of the SBE stack; evidence is on #123.

[bfrancojr]

Rebased with the stack onto main df6b102; head is now db34ed3. Evidence is on #124.

[bfrancojr]

Rebased onto main 197e05b, after #124 and #137 merged as squashes: db34ed3 → 39e1261. The old #124 and #137 commits were dropped (git rebase --onto origin/main bcedc19), so this PR's own commit is the only one replayed. Its diff is unchanged, apart from where its CHANGELOG line lands. At the stack top (39e1261): build OK, ctest 100% tests passed out of 705. The remote head was read back with git ls-remote: 39e1261.

[bfrancojr]

Rebased onto main 9525e9b, after #140 merged as a squash: 39e1261 → 2ee007c. I dropped #140's old commit with git rebase --onto origin/main 3486d29. The diff is unchanged apart from where the CHANGELOG line lands. At the stack top (2ee007c): build OK, ctest 100% tests passed out of 705. The remote head was read back with git ls-remote: 2ee007c.

[bfrancojr]

Rebased onto main 1706247, after #143 merged as a squash: 2ee007c → 674b935. I dropped #143's old commit with git rebase --onto origin/main 72c54f1. The diff is unchanged, and it still adds one CHANGELOG line. Build OK, ctest 100% tests passed out of 705. The remote head was read back with git ls-remote: 674b935.