Winarc (wep32-community)

Borland OWL puzzle collection: binaries/wep32-community/Winarc/Winarc.exe, registered as winarc. The 400x300 launcher holds only a menu; every puzzle (Bishops, Boxes, Hex, Knights, Krypto, Life, Loyd, Mazes, Memory, Pegs, Queens, Rubik, Sieve, Spaceship, Sudoku, Triads) is its own top-level OWL_Window created from the Window menu. A few (help.txt/noname.txt RICHEDIT viewers, the "colors" dialog) are created hidden at startup.

Route (headless)

Menus open with press, gap, release: 1500:mousedown:39:31,1520:mouseup:39:31 opens Window; 1570:mousemove:42:151,1590:mousedown:42:151,1610:mouseup:42:151 picks Life (Queens is at 52,251). F5 (keydown:0x74/keyup) is Step.

Life draws slowly -- capture after the first paint

The Life window ("noname.lif", title generation N [cells; 51x58 toral]) paints its board in one WM_PAINT: CreateCompatibleBitmap(312x274), then the bitmap is selected into three successive memory DCs (deselected before each DeleteDC), each live cell is one Rectangle, and the last DC is BitBlt'd to the BeginPaint DC. That one paint is ~3,500 API calls, so at the default batch size the board first appears around batch 2,900 although the title already reads generation 0 by batch ~2,200. A capture in between shows a blank grey client and looks like a lost blit; it is not (memory-DC reuse is correct). Each F5 is ~3,700-4,100 calls and two board blits. Evidence: scratch/runs/20261010T1100Z-winarc-life-step-w6.

Open

• The menu accelerator labels read ?F5: the strings are built at runtime (no F5 text in the resources); not investigated.
• Queens draws its board the same way (240 Rectangles) but has not been taken through a reviewed A/B.