Watch the run graph, and say what was refused #296

Merged

PerishFire merged 1 commit from land/task/watch-the-run-graph into main

2026-08-18 05:09:14 +00:00

PerishFire commented

2026-08-18 05:06:09 +00:00

Owner

Watch the run graph, and say what was refused

A dispatch's own success is only news that the trigger landed, and the watch
that followed it was nearly as forgiving: a run reporting success while its job
list was empty read as finished, and a run reporting success while one of its
jobs had failed read as finished too. Both are how a release comes back green
and publishes nothing, and both have been misread from this repository more than
once.

The verdict now comes from the graph. No jobs yet means the run has not laid
itself out, so keep waiting; a failed job means failure whatever the run says;
success needs every job terminal and none failed. That verdict is a plain
function of the task list, so it is tested as one.

Two refusals also stopped speaking in riddles. A forge error now names the
operation Plumb was performing when it arrived, instead of arriving as a bare
Please try again later with no subject; the flags are dropped from that name
so no argument value rides along. And head out of date, which Forgejo says to
any head it does not like, now names the commit the pull actually stands at — or
says the pull does stand there and the refusal was about something else.

Co-Authored-By: Claude Opus 5 (1M context) noreply@anthropic.com

Watch the run graph, and say what was refused A dispatch's own success is only news that the trigger landed, and the watch that followed it was nearly as forgiving: a run reporting success while its job list was empty read as finished, and a run reporting success while one of its jobs had failed read as finished too. Both are how a release comes back green and publishes nothing, and both have been misread from this repository more than once. The verdict now comes from the graph. No jobs yet means the run has not laid itself out, so keep waiting; a failed job means failure whatever the run says; success needs every job terminal and none failed. That verdict is a plain function of the task list, so it is tested as one. Two refusals also stopped speaking in riddles. A forge error now names the operation Plumb was performing when it arrived, instead of arriving as a bare `Please try again later` with no subject; the flags are dropped from that name so no argument value rides along. And `head out of date`, which Forgejo says to any head it does not like, now names the commit the pull actually stands at — or says the pull does stand there and the refusal was about something else. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

PerishFire added 1 commit

2026-08-18 05:06:09 +00:00

Watch the run graph, and say what was refused

guard / guard (pull_request) Successful in 2m53s

Details

guard / guard (push) Successful in 2m43s

Details

da6cf55736

Watch the run graph, and say what was refused

A dispatch's own success is only news that the trigger landed, and the watch
that followed it was nearly as forgiving: a run reporting success while its job
list was empty read as finished, and a run reporting success while one of its
jobs had failed read as finished too. Both are how a release comes back green
and publishes nothing, and both have been misread from this repository more than
once.

The verdict now comes from the graph. No jobs yet means the run has not laid
itself out, so keep waiting; a failed job means failure whatever the run says;
success needs every job terminal and none failed. That verdict is a plain
function of the task list, so it is tested as one.

Two refusals also stopped speaking in riddles. A forge error now names the
operation Plumb was performing when it arrived, instead of arriving as a bare
`Please try again later` with no subject; the flags are dropped from that name
so no argument value rides along. And `head out of date`, which Forgejo says to
any head it does not like, now names the commit the pull actually stands at — or
says the pull does stand there and the refusal was about something else.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Land-Source: task/watch-the-run-graph@857fd2c223

PerishFire merged commit da6cf55736 into main

2026-08-18 05:09:14 +00:00

PerishFire referenced this pull request

2026-08-18 05:26:03 +00:00

Release failure issues must capture the failing critical step #136