Repository navigation
Commit b5d4ea0
authored
ci: measure coverage, and drop tests other tests already cover (#13)
* ci: measure test coverage of the hooks
claude plugin test measures no coverage, and the hooks run in the
engine's own environment, where no coverage tool sees them. So
npm run coverage instruments a copy of the plugin with istanbul,
has each test read the counts back through a __coverage command when it
ends, and reports them as text and lcov. CI prints the table in the
job summary.
* test: drop the tests other tests already cover
Coverage per test showed 11 tests whose lines the rest run too: the
pane listing per surface, the session press, the full replay, the
kept-token and CLAUDE_CONFIG_DIR connects, /leverage with no token,
the subagent step, the network retry and logout. Line coverage goes
from 62.18% to 61.69%.
The token-safety tests stay though others run their lines: they are
the only ones that check a token is never kept where others can read
it.
* ci: comment the coverage table on each PR
One comment per PR, edited on each push. Skipped for forks, whose
token cannot comment.1 parent 73ec5f9 commit b5d4ea0
8 files changed
Lines changed: 2917 additions & 241 deletions
File tree
- .github/workflows
- scripts
- tests
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
6 | 6 | | |
7 | 7 | | |
8 | 8 | | |
| 9 | + | |
9 | 10 | | |
10 | 11 | | |
11 | 12 | | |
| |||
23 | 24 | | |
24 | 25 | | |
25 | 26 | | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
| 30 | + | |
| 31 | + | |
| 32 | + | |
| 33 | + | |
| 34 | + | |
| 35 | + | |
| 36 | + | |
| 37 | + | |
| 38 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
2 | 2 | | |
3 | 3 | | |
4 | 4 | | |
| 5 | + | |
0 commit comments