diff options
Diffstat (limited to 'features.txt')
| -rw-r--r-- | features.txt | 11 |
1 files changed, 11 insertions, 0 deletions
diff --git a/features.txt b/features.txt index 5583e7f3..eeadba39 100644 --- a/features.txt +++ b/features.txt @@ -156,3 +156,14 @@ when it is false and nothing is animating. 9P round trip on a local socket: gui the problem -- instrumentation proved every request woke the loop early (wakes=210, woke_early=212, timed_out=38 over 250 ticks) -- the reply simply could not be produced until the loop had finished drawing a frame it did not need. Verified the gui still paints (screen shows the file, and a write through 9P redraws) and the full suite is unchanged at 778/783 with the two known crashes. + +Found by adversarial review and NOT fixed, because they are upstream or macOS-only rather than from this session's work: +- Surface.mark_hover is never assigned true anywhere in the tree, so the whole macOS hover ABI is inert: PARDES_CELL_HOVER is never emitted, encodeCellFlags always + contributes 0, and PardesView's liquid-glass affordance never draws. The tests pass only because they set the flag by hand. The comment in the look_hover_preview + paint block claims it carries those cells out to the hosts; it does not. +- -Dworkspace-tag is read only in src/macos.zig, and core.settings.workspace_tag is assigned only there, so the option builds cleanly for gui/tty/web and is + silently ignored. gui.zig hard-codes `true` at six taglineBandOffset call sites and still uses raw TOPBAR_H at five more. +- The detached wire encodes the pointer shape in the frame header variant (full_link/diff_link), so the newer `.target` shape collapses to `.arrow` for an attached + frontend. Fixing it means a new header variant, i.e. a protocol change. +- A shell that outlives its editor keeps PARDES_PID; if that pid is reused, `pardes <file>` now exits 1 instead of falling back to starting an editor. Exiting 1 + only when the socket itself refused would keep the old behaviour. |