PCFC11 0.2.27 — Windows x64 validation

1. Extract PCFC11-0.2.27-source.zip into a NEW folder. Keep every folder together.
   Reuse your native Windows Python 3.11+ and MinGW-w64 UCRT64 GCC installation.
   PyCForge does not need to be installed for this standalone test.

2. Open PowerShell in the extracted PCFC11-0.2.27 directory and run:

   python tools/test_windows.py

   Or open MSYS2 UCRT64 / Git Bash and run:

   bash test-pcfc11.sh

   If GCC is outside PATH:

   python tools/test_windows.py --cc "C:\msys64\ucrt64\bin\gcc.exe"

3. Allow several minutes. The runner builds pcfc11.exe and prints each phase.
   Phases test 146 initializer cases and 130 bit-field layouts, repeated header inclusion, standard headers and 49 generated programs at
   PCFC11 O0/O1 and GCC O0/O2. Python supplies output/file references; one
   explicitly labelled case checks PyCForge's signed-64 overflow policy.
   Existing language, ABI, optimizer, robustness and Unicode checks remain.
   If a phase times out, add --timeout 1200 to allow 20 minutes per phase.

4. Return the printed test-results\windows-*.zip even if it says FAIL.
   Include which terminal you used and any Windows blocking/manual action.
   If setup fails before a ZIP exists, return the printed setup-*.txt instead.

Nothing is installed or uploaded automatically. Reports can contain local paths.
GCC supplies assembly, linking and UCRT; PCFC11 supplies C compilation.
Keep include/ beside the compiler. MSVC-hosted builds are not supported.
WSL validates Linux, not Windows. macOS comes later.

The native 0.2.5 run passed 16/17 phases and all 49 PyCForge cases. Its assertion
header failure exposed an include-guard defect, fixed in 0.2.6. The supplied 0.2.7
native run passes all 19 phases. Version 0.2.8 added bit-field support. The supplied 0.2.9 run passes all
20 phases, including scalar initialization and the protected-page checks.
Bit-field skips are removed. VLAs and some alignment cases remain unsupported.
The runtime headers are a partial bridge, not a complete Windows SDK or C11
library. A pass does not certify unwind/guard-page handling or complete C11.
These tests use pinned generated C; live Windows PyCForge GUI/build-service
validation remains separate. PyCForge 1.3.57 and its compiler choices are unchanged.
The report records the installed PyCForge distribution version visible to this
Python interpreter. That is metadata, not a live conversion test; the bundled
C fixture version stays at 1.3.57 regardless of the installed application.

Optional same-host timing comparison:

   python tools/test_windows.py --baseline "C:\path\to\PCFC11-0.2.19\pcfc11.exe"

Or in Bash:

   bash test-pcfc11.sh --baseline 'C:/path/to/PCFC11-0.2.19/pcfc11.exe'

Keep the previous compiler and its own include/ folder together. Use the actual
path to your retained 0.2.19 build; omit --baseline if you do not have it.

Version 0.2.27 corrects C11 generic-selection diagnostics, variable-length
array control conversion, nested association restrictions and fixed/VLA type
compatibility. The constraint matrix contains 350 cases. Twenty-six new VLA
cases exercise the existing Windows unsupported-feature diagnostic, not native
VLA execution. The standard runner still has 36 native Windows phases.
Native Windows 0.2.27 execution is pending. The earlier 36-phase 0.2.25 feedback
is retained with its original scope. Return the new feedback ZIP even on failure.
See docs/GENERIC_SELECTION.md and docs/VALIDATION.md for details.
