2.7 KiB
libs/libcmt/putws.obj provenance park
This isolated lane starts from authoritative campaign commit 3d2b81dd.
Configuration index 719 remains MISSING. The unit was unparked, unowned,
and absent from prior source/history/logs. No production or configuration file
was edited.
Plain-English behavior and provenance
_putws locks standard output, temporarily prepares its buffer, writes every
wide character followed by a newline, restores buffering, and always unlocks
the stream through a Microsoft SEH finally cleanup. It returns zero on
success and wide EOF on failure.
The exact obj\i386\putws.obj member was extracted from XDK 3911
libcmt.lib. CodeView identifies ordinary Microsoft C at
d:\xbox-finalfre\private\sdktools\crt\crtw32\stdio\puts.c, compiled as its
wide specialization. This is not assembly.
Measured blocker
January and the authentic member both contain a 143-byte function and the same 12-byte SEH scope table. The authentic compiler member has 11 text relocations; January's split representation has 12. Two sites prove that the difference is extraction metadata, not source shape:
| Site | January split identity | Authentic XDK identity |
|---|---|---|
+0x72 |
synthetic internal REL32 to cleanup at +0x84 |
encoded local displacement +0x0E, no relocation |
+0x88 |
REL32 labeled _ai_debug + 216415 |
REL32 to the real callee __unlock_file2 |
The second identity is especially decisive: the disassembly and Microsoft
source both show _unlock_file2(stdout). No honest reconstruction may replace
that call with an unrelated Halo AI routine merely to satisfy a splitter's
symbol guess. The target's effective linked instruction behavior corresponds
to the authentic member, but strict COFF relocation identity correctly refuses
the extracted representation.
Decision and reopen criterion
No source experiment was run. Production/configuration remain pristine. This
is a docs-only park because source cannot manufacture the synthetic local
relocation or the incorrect _ai_debug identity without byte forcing and
semantic corruption.
Reopen only after a reviewed target-extraction fix restores _unlock_file2
and preserves the compiler's local finally displacement, or after a
fail-closed comparator adjudication specifically proves these csplit artifacts.
House, Berth, bug, and integrity audit
No assembly, volatile, forced inline, undefined behavior, flag change, byte
patch, comparator exception, or prebuilt substitution was introduced. No
original implementation bug was found; the SEH cleanup is intentional. The
only repository change is this plain-English evidence log. Tooling tests passed
179/179 and git diff --check is clean.