Environment
| Item | Value |
|---|---|
| Product | Google Antigravity (Electron hub) |
| App version (logs) | 2.28.1 |
| Package | Arch AUR antigravity 2.28.1-1 (official linux-x64 tarball from storage.googleapis.com/antigravity-public/...) |
| Build id (PKGBUILD) | 6709500728705024 |
| OS | Omarchy 4.0.6 (Arch-based) |
| Kernel | 7.2.8-9-omarchy x86_64 |
| Desktop | Hyprland 0.56.2, Wayland |
| GPU | Intel UHD (Comet Lake) + NVIDIA GA106M (RTX 3060 Mobile) |
| Install path | /opt/Antigravity/antigravity |
Summary
On first launch after installing 2.28.1, the main Antigravity process aborted with SIGTRAP (int3 / ud2) during shutdown. The app had already logged Shutting down language server… a few seconds earlier. This looks like a Chromium/Electron FATAL/CHECK abort on the teardown path, not an OOM kill and not a packaging crash.
Related prior reports of exit-time SIGTRAP / coredumps on Linux (utility-process variants):
This report is for the main process (PID matching the app scope), not node.mojom.NodeService.
Timeline (EDT, 2026-10-31)
- 21:44:23 — package installed (
pacman -Uof AUR build wrapping the official tarball). - 21:44:29 — first start;
~/.config/Antigravity/logs/main.logshows normal startup (Starting app (v2.28.1)…, host bridge, AutoUpdater). - Ran ~1m 43s wall clock, peak RSS ~389 MiB.
- 21:46:07 —
main.log:Shutting down language server… - 21:46:11 — kernel:
traps: antigravity[…] trap int3→ systemd-coredump signal 5 (TRAP).
coredumpctl recorded the command line as /opt/Antigravity/antigravity --version. On this build, --version still starts the full GUI (confirmed on a later launch); the process was a normal first-run session, not a print-and-exit.
Evidence
Not OOM
- No
systemd-oomd/ “Out of memory” / “Killed process” journal entries around the crash. - Available memory was fine at investigation time; crash peak ~389 MiB.
Kernel / coredump
traps: antigravity[152302] trap int3 ip:… in antigravity[…]
Signal: 5 (TRAP) si_code: SI_KERNEL
Executable: /opt/Antigravity/antigravity
GDB on the core (binary stripped; no public debuginfod symbols) shows the classic Chromium fatal pattern on the main thread:
call <…>
int3
ud2 ; RIP here
Main-thread shape (unsymbolized): Antigravity frames → g_main_context_iteration (libglib) → Antigravity → __libc_start_main. Other threads were idle (cond wait / epoll / glycin). Crashpad under ~/.config/Antigravity/Crashpad/ had only a client_id (no usable .dmp with a CHECK message).
App log (crashed session)
[2026-10-31 21:44:29.120] Starting app (v2.28.1) with dynamic port…
…
[2026-10-31 21:44:39.488] [AutoUpdater] Starting auto update checks
[2026-10-31 21:44:39.489] APPIMAGE env is not defined, current application is not an AppImage
[2026-10-31 21:46:07.106] Shutting down language server…
(No FATAL line in main.log before the trap.)
Expected
Clean exit after shutdown without SIGTRAP / systemd-coredump.
Actual
Main process dumps core with SIGTRAP during/immediately after language-server shutdown.
Repro notes
- Observed once on first run after install; a subsequent short launch did not reproduce.
- Likely trigger: normal quit / teardown after a brief first session on Wayland + Hyprland.
- Binary is stripped, so stacks stay unresolved without Google-side symbols.
Happy to capture a fresh core + main.log / language_server.log if engineering wants a specific repro recipe (e.g. --enable-logging, Crashpad forced dump).