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
opencode 1.18.21 crashes with SIGILL at UD2 instruction #46172
The core contained 29 threads. Most other threads were in runtime/internal wait or worker code. The coredump also recorded two deleted temporary modules without build IDs:
There was no evidence of OOM: at diagnosis time the system had approximately 6.7 GiB available memory and no swap in use.
This appears to be an internal fatal path or assertion in the opencode executable, but the available symbols do not identify the triggering code path.
The crash has only been observed once so far.
Expected behavior
The opencode process should continue running or report a recoverable application error rather than terminate with SIGILL.
Additional information
The coredump is available locally through systemd-coredump, but it has not been uploaded because it may contain private prompts, tokens, and user data.
Filed by the current model via pi coding-agent harness.
This issue might be a duplicate of existing issues. Please check:
TUI SIGILL (ud2) in OpenTUI drawTextBuffer when compositor scale jumps to 4 #42094: Same ud2 abort at identical offset 0x2607064, same ILL_ILLOPN signal code, same Arch Linux + systemd-coredump setup, and the same deleted /tmp/*.so pattern. That report identifies the crash site as a Zig/Bun panic in OpenTUI's drawTextBuffer path.
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.
Activity
github-actions commentedon Aug 29, 2026
This issue might be a duplicate of existing issues. Please check:
ud2abort at identical offset0x2607064, sameILL_ILLOPNsignal code, same Arch Linux + systemd-coredump setup, and the same deleted/tmp/*.sopattern. That report identifies the crash site as a Zig/Bun panic in OpenTUI'sdrawTextBufferpath.ip:2607064) inside the opencode binary on an Intel CPU running a Linux distro with systemd-coredump. Discusses possible CPU micro-arch baseline mismatch.ud2/libopentui+0x1d061dabort site.The crash address
0x2607064appears consistently across these reports, suggesting they share a root cause.thdxr commentedon Oct 11, 2026
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (
opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊