diag: surface odd-length hex output from MEOS *_as_hexwkb on macOS arm64 (refs #126/#130)

[estebanzimanyi]

Summary

Stacks on #158 (the edge-to-cloud quickstart substrate). Wraps every *_as_hexwkb call-site in MobilityDuck with a diagnostic-grade precondition check that surfaces odd-length hex output at the producer rather than the consumer.

What this is for

PRs #126 and #130 have been failing the macOS osx_arm64 unittest with:

Invalid hex string, length (69) has to be a multiple of two!

That message is from DuckDB's hex/blob decoder rejecting an odd-length string — meaning one of the MEOS *_as_hexwkb producers is emitting a 69-char hex string on that platform. Hex strings of WKB-encoded MEOS values must always be even-length (each byte is encoded as 2 hex chars), so the producer is either:

• producing a literal odd-length string (MEOS-side bug), or
• producing an even-length string with an embedded null at an odd position (DuckDB sees the null as the string end via strlen)

Instrumentation pattern

At every *_as_hexwkb call-site, immediately after the MEOS call returns:

size_t hex_size = 0;
char *hex = X_as_hexwkb(..., &hex_size);
if (!hex) throw InternalException("asHexWKB: X_as_hexwkb failed");

// Diagnostic: hex strings must be even-length.
size_t actual = strlen(hex);
if (actual % 2 != 0) {
    std::string diag = "asHexWKB: X_as_hexwkb produced odd-length string: "
                       "strlen=" + std::to_string(actual) +
                       " sz=" + std::to_string(hex_size);
    free(hex);
    throw InternalException(diag);
}
string_t stored = StringVector::AddString(result, hex, actual);  // explicit length
free(hex);

When the producer emits odd-length output on macOS-arm64, the exception message includes both strlen(hex) and the sz out-parameter reported by MEOS — pinpointing which side is wrong + what the truncation point was.

The explicit-length StringVector::AddString(result, hex, actual) (instead of the implicit strlen overload) also makes any latent embedded-null bug detectable at the producer rather than the consumer.

Sites instrumented (7 producers)

File	Function	MEOS export
src/geo/tgeompoint.cpp	TgeoAsHexWkbExec	temporal_as_hexwkb
src/geo/stbox_functions.cpp	Stbox_as_hexwkb	stbox_as_hexwkb
src/temporal/span_functions.cpp	Span_as_hexwkb	span_as_hexwkb
src/temporal/spanset_functions.cpp	Spanset_as_hexwkb	spanset_as_hexwkb
src/temporal/set_functions.cpp	Set_as_hexwkb	set_as_hexwkb
src/temporal/tbox_functions.cpp	Tbox_as_hexwkb	tbox_as_hexwkb
src/temporal/temporal_functions.cpp	Temporal_as_hexwkb	temporal_as_hexwkb

No-op on green paths

The change is intentionally a no-op on green platforms: hex strings that are already even-length are passed through with no behaviour change, just an explicit length argument. The diagnostic path only fires when the bug is present.

Local verification

Linux x86_64 — full release build green (498/498 ninja targets):

$ ./build/release/duckdb -c "LOAD '...mobilityduck.duckdb_extension'; SELECT asHexWKB(tgeompoint 'Point(1 1)@2020-01-01');"
012E00010101000000000000000000F03F000000000000F03F00BC2EB0063E0200

Even-length hex output; diagnostic does not trigger on the happy path.

Branch base

feat/edge-to-cloud-quickstart-rebased (= #158's branch). Polyfills (#136 ICU staging, #140 bigint_to_set forwarder, #173 unique_ptr fix) inherited from the substrate stack.

Refs

• feat(parity): batch — bearing + eCovers/tCovers + stbox dim + seqSetGaps (consolidates #122+#123+#124+#125) #126, feat(parity): close final 14-name gap to reach 100% addressable coverage #130 — macOS osx_arm64 unittest failures the diagnostic is for
• feat: edge-to-cloud quickstart, temporalFooter, SRID/geodetic fix, tgeogpoint tests (supersedes #113) #158 — substrate this stacks on
• [[mobilityduck-hex-instrumentation-draft]] — the design memo