• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

Tatsh / sbclaude / 32911047693
100%

Build:
DEFAULT BRANCH: master
Ran 25 Aug 2026 11:30PM UTC
Jobs 5
Files 5
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

25 Aug 2026 11:29PM UTC coverage: 100.0%. Remained the same
32911047693

push

github

Tatsh
project: retry a box that never starts, and say why it did not

A box that dies before the session begins left nothing to go on. The
fullscreen TUI puts claude on the terminal's alternate screen and that
screen is discarded on exit, so whatever was printed on the way down went
with it; docker's own diagnosis on stderr survived, but the `docker run`
behind it -- which is usually where the answer is -- was never shown at
all.

Three things, all about that moment:

Retry. Exit codes 125-127 are what the docker CLI reserves for a
container that never ran, and a box that exits within seconds died before
a session could begin. Either is a failure to start and is launched
again, three attempts in all. A signal death (128+) is neither -- a
session ended with Ctrl-C did exactly what was asked -- and a session
that ran and then ended is never retried, whatever its exit code. The
command is rebuilt per attempt rather than reused, which is what lets the
GPU probe below take effect on the next try.

Somewhere to read it afterwards. Every failed attempt goes to syslog with
docker's message and the full command, one entry per line because a
single long entry is what journalctl truncates at the terminal width. A
syslog that cannot be written says so and does not take the session down
with it. `docker run` is now spawned through Popen so its stderr can be
relayed as it arrives and a bounded tail kept back for the log; stdin and
stdout stay on the terminal untouched, because -t and the TUI both need a
real one and the container's output already comes back on stdout.

--no-fullscreen (config key fullscreen) leaves the TUI setting as the
host has it, so a failing session's output stays on the normal screen.
The failure message points at it, but only when the box got far enough
that claude could have printed something.

And the cause that prompted all this: --gpus all with no usable card is
unresolvable against the CDI registry, which the daemon treats as a
warning rather... (continued)

224 of 224 branches covered (100.0%)

Branch coverage included in aggregate %.

101 of 101 new or added lines in 3 files covered. (100.0%)

863 of 863 relevant lines covered (100.0%)

5.0 hits per line

Jobs
ID Job ID Ran Files Coverage
1 32911047693.1 25 Aug 2026 11:30PM UTC 5
100.0
GitHub Action Run
2 32911047693.2 25 Aug 2026 11:30PM UTC 5
100.0
GitHub Action Run
3 32911047693.3 25 Aug 2026 11:30PM UTC 5
100.0
GitHub Action Run
4 32911047693.4 25 Aug 2026 11:30PM UTC 5
100.0
GitHub Action Run
5 32911047693.5 25 Aug 2026 11:30PM UTC 5
100.0
GitHub Action Run
Source Files on build 32911047693
  • Tree
  • List 5
  • Changed 3
  • Source Changed 3
  • Coverage Changed 3
Coverage ∆ File Lines Relevant Covered Missed Hits/Line Branch Hits Branch Misses
  • Back to Repo
  • Github Actions Build #32911047693
  • 27f0382f on github
  • Prev Build on master (#32476615199)
  • Next Build on master (#32930334449)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc