You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
macOS 27: read push fails (exit 75) when Messages' window is minimized or closed #164
imsg-read --all exits 75 whenever Messages' main window is minimized, or when Messages is running with no window:
imsg-read: 1 unread on the Mac but Mark All as Read stayed disabled even after waking Messages, so the click could not land
wake_messages() activates Messages, which does make it frontmost, but activation does not deminiaturize a minimized window or open a closed one. With no visible main window, Conversation ▸ Mark All as Read stays disabled even while Messages is the active app.
On my gateway this failed silently from Oct 20 to Nov 11. push-read.log has ~20 consecutive exit-75 entries with one success, and Blip's local read state hid it.
Same Mac, same unread message, same ssh grants. Only the window state changes:
Messages window
imsg-read --all
Dock badge after
minimized
exit 75 (above)
1
restored
exit 0, marked (0 unread)
none
Also observed:
No window at all (Messages running, count windows = 0): Mark All as Read stayed enabled=false at +0.7 s, +2 s and +4 s after activate, with Messages frontmost the whole time.
Not locked (CGSSessionScreenIsLocked absent), Stage Manager off, default Dock minimize settings.
blip-check: chatdb, automation and contacts all pass. imsg-read --status: ready.
Not a permissions problem
The Accessibility grant imsg-read already holds is enough to fix it. Over the same ssh session:
tell application "System Events" to set value of attribute "AXMinimized" of window 1 of process "Messages" to false
restores the window, and the push then succeeds.
Suggested fix
In the dormant-menu path, before clicking:
if Messages' window is minimized, restore it (AXMinimized false)
if Messages has no window, reopen it
Then put it back afterwards, the way restore_front() hands focus back. If that's unwanted, a README line under "Keeping the Mac awake" saying "keep a Messages window open and not minimized on the gateway" would at least document it, and --status could detect the case and report it instead of only "dormant".
Unverified: whether the same happens on macOS 26. The 0.7 s activation note in CLAUDE.md was measured on 26.6, possibly with a visible window.
Thank you Jared, this is a clear report and the repro table is exactly what was needed. It is confirmed as a real gap (Mark All as Read needs a window on screen, and activation does not restore a minimized one or open a closed one). A fix along your suggestion (restore the minimized window, or reopen when there is none, then put it back) is in progress and not committed yet; I will report here when it lands. It should compose with #167, whose lock-screen check would stop the reopen from running behind a locked screen. The README line about keeping a Messages window open is a good fallback and will go in either way.
Fixed in e76abcd, thank you Jared: your repro table and the AXMinimized suggestion were exactly right. imsg-read now reads the Messages window state first (read-only), restores a minimized window or asks Messages to reopen when there is none, clicks Mark All as Read, then minimizes it again (a window it had to open is left open; the window goes back even if the click fails; an unreadable state changes nothing). imsg-read --status also names the window as the reason when the menu is dormant. Tests: 19 pass, on a Python 3.9.6 gateway too, and each guard breaks a test when removed. On a real 27.2 gateway I confirmed the window-state read and the minimize/restore with the tool's own functions; I have not yet run a full mark-all with a minimized window and an unread message, so tell me if it still exits 75 for you. Your lock-screen guard in #167 fits next to this (it stops the reopen from running behind a locked screen).
Activity
nixfred commentedon Nov 13, 2026
Thank you Jared, this is a clear report and the repro table is exactly what was needed. It is confirmed as a real gap (Mark All as Read needs a window on screen, and activation does not restore a minimized one or open a closed one). A fix along your suggestion (restore the minimized window, or reopen when there is none, then put it back) is in progress and not committed yet; I will report here when it lands. It should compose with #167, whose lock-screen check would stop the reopen from running behind a locked screen. The README line about keeping a Messages window open is a good fallback and will go in either way.
fix(bridge): Mark All as Read no longer fails with exit 75 when Messa…
nixfred commentedon Nov 14, 2026
Fixed in e76abcd, thank you Jared: your repro table and the AXMinimized suggestion were exactly right.
imsg-readnow reads the Messages window state first (read-only), restores a minimized window or asks Messages to reopen when there is none, clicks Mark All as Read, then minimizes it again (a window it had to open is left open; the window goes back even if the click fails; an unreadable state changes nothing).imsg-read --statusalso names the window as the reason when the menu is dormant. Tests: 19 pass, on a Python 3.9.6 gateway too, and each guard breaks a test when removed. On a real 27.2 gateway I confirmed the window-state read and the minimize/restore with the tool's own functions; I have not yet run a full mark-all with a minimized window and an unread message, so tell me if it still exits 75 for you. Your lock-screen guard in #167 fits next to this (it stops the reopen from running behind a locked screen).