kyle/crossbar
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

ASSESSMENT: the idle-session cause was OpenCode 1.15 waiting on a piped stdin

Browse Source 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
kyle
2026-09-25 22:39:02 -07:00
co-authored by Claude Opus 5.5
parent c9c9af908d
commit 2683df51bd
1 changed files with 7 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ASSESSMENT.md
+7 -3
View File
@@ -188,9 +188,13 @@ Ornith's time was lost mostly to the owner's tests and task text. By category:
- **The OpenCode sandbox** refuses anything outside the repository, including `/tmp` and the nix - **The OpenCode sandbox** refuses anything outside the repository, including `/tmp` and the nix
store. This is the single biggest source of early endings; design tasks so nothing outside the store. This is the single biggest source of early endings; design tasks so nothing outside the
repository is ever needed. repository is ever needed.
- **OpenCode upgrades** ran a one-time database migration on the next start. One session then - **OpenCode 1.15 hangs as a background job unless stdin is closed.** After an upgrade (it ran a
sat idle for 46 minutes with no model request, all llama slots idle. A trivial prompt worked one-time database migration), `opencode run` started reading a piped stdin as extra prompt
afterwards. The driver had no timeout, so nothing noticed. and waiting for EOF. A backgrounded driver's stdin never closes, so two resume sessions sat
idle for 46 and 10 minutes without sending one model request (all llama slots idle), while
the same command in the foreground worked. Confirmed side by side: exit 124 without, `pong`
with `</dev/null`. Both `run-plan.sh` scripts now pass `</dev/null`. The driver had no timeout,
so nothing noticed.
## 8. Recommendations ## 8. Recommendations