Raw evidence file inlined. Source path on boltzmann: /home/mfritsche/src/kwin_overlay_subsurface/metrics.csv.

# kwin_overlay_subsurface metrics — locked Phase 1 cells
#
# Inherits the success-criteria shape from ohm_gl_fix Phase 1r
# (drops_post_warmup, drops sanity cap) plus a new cell
# kwin_wayland_cpu_steady_state that locks the intent that the
# fix must move work, not hide it.
#
# Schema:
#   phase             phase0_handover | phase1_baseline_kwin
#                     | phase1_reference_cage | phase1_goal_target
#                     | phase1_goal_actual | phase3_perf_kwin
#                     | phase3_perf_cage
#   path_label        descriptive playback configuration
#   clip              source media file (sha16: dcf8a7170fbd49bb for bbb_1080p30_h264.mp4)
#   compositor        kwin | cage_nested | cage_drm
#   surface_protocol  zwp_linux_dmabuf_v1_subsurface | single_fullscreen_surface | direct_drm_scanout
#   drops             total dropped frames over window_s
#   frames_total      total frames considered (delivered + dropped)
#   drop_pct          100 * drops / frames_total
#   window_s          full measurement window
#   warmup_s          duration tolerated as warmup
#   drops_post_warmup drops in (warmup_s, window_s] — binding for C1
#   effective_fps     delivered frames per second
#   kwin_wayland_cpu  steady-state %CPU of kwin_wayland during t > 30 s
#   route_engaged     Y/N — wp_subsurface protocol still in use (binding for C4)
#   role              metric=success criterion;
#                     reference=control floor;
#                     metric_actual=measured-against-locked-metric;
#                     handover=imported context, not own measurement
#   source            where this row's number came from
#   date              ISO date the number was taken
#
# Phase 1 success criteria (locked 2026-05-02):
#   C1: drops_post_warmup == 0 over 60 s (warmup_s = 10)
#   C2: drops_total <= 10 over 70 s (sanity cap)
#   C3: kwin_wayland_cpu <= cage_baseline + 5pp absolute
#   C4: route_engaged == Y (no "treat-as-regular-surface" cheat)
# All four must hold for Phase 7 verification.
phase,path_label,clip,compositor,surface_protocol,drops,frames_total,drop_pct,window_s,warmup_s,drops_post_warmup,effective_fps,kwin_wayland_cpu,route_engaged,role,source,date
phase0_handover_kwin,brave_chromium_step1_step2_qt6kwin_fourier,bbb_1080p30_h264.mp4,kwin,zwp_linux_dmabuf_v1_subsurface,58,1498,3.87,70,10,29,21.3,20.5,Y,handover,ohm_gl_fix:phase3_remeasure_2026-05-02/A2_brave_drops_findings.md,2026-05-02
phase0_handover_cage,brave_chromium_step1_step2_qt6kwin_fourier,bbb_1080p30_h264.mp4,cage_nested,single_fullscreen_surface_to_kwin,7,1686,0.42,70,10,0,24.0,,Y,handover,ohm_gl_fix:phase3_remeasure_2026-05-02/task25_cage_vs_kwin_decisive.md,2026-05-02
phase1_goal_target,brave_under_kwin_with_patch,bbb_1080p30_h264.mp4,kwin,zwp_linux_dmabuf_v1_subsurface,10,1440,0.69,70,10,0,24.0,cage_baseline_plus_5pp,Y,metric,kwin_overlay_subsurface:phase1_lock.md,2026-05-02
phase3_perf_kwin,chromium_fourier_step1_step2_qt6kwin_fourier_balooOff,bbb_1080p30_h264.mp4,kwin,zwp_linux_dmabuf_v1_subsurface,TBD,TBD,TBD,70,10,TBD,TBD,17.0,TBD,metric_actual,kwin_overlay_subsurface:phase3_findings.md (drops trajectory not captured; see phase3_evidence/kwin_direct_rep1/),2026-05-02
phase3_perf_cage,chromium_fourier_step1_step2_qt6kwin_fourier_balooOff,bbb_1080p30_h264.mp4,cage_nested,single_fullscreen_surface_to_kwin,TBD,TBD,TBD,70,10,TBD,TBD,36.9,TBD,metric_actual,kwin_overlay_subsurface:phase3_findings.md (kwin_wayland_cpu measured on KWin under cage; cage process perf also captured at ~7%; drops trajectory not captured),2026-05-02
phase3_perf_brave_kwin,brave_stock_v147_no_fourier_balooOff,bbb_1080p30_h264.mp4,kwin,unknown_no_chrome_trace,18,1624,1.11,70.27,10,0_visible_upper_bound_18,23.1,35.9,unknown_no_chrome_trace,metric_actual_exploratory,kwin_overlay_subsurface:phase3_findings.md C3 + phase3_evidence/brave_stock_kwin_rep1/drops_trajectory_typed.txt,2026-05-02