diff options
| author | Craig Jennings <c@cjennings.net> | 2026-05-14 07:51:02 -0500 |
|---|---|---|
| committer | Craig Jennings <c@cjennings.net> | 2026-05-14 07:51:02 -0500 |
| commit | e98d9b02a167fa3db19dc7f459b698df3c619b9a (patch) | |
| tree | c1eeda3f883d2e188aad41061da7f45813a09107 /tests/test-setup-telega.bats | |
| parent | 4706cd0a46e6274e512c159d0d6ec71a3a47a20f (diff) | |
| download | dotemacs-e98d9b02a167fa3db19dc7f459b698df3c619b9a.tar.gz dotemacs-e98d9b02a167fa3db19dc7f459b698df3c619b9a.zip | |
test(system-defaults): switch to single top-level require so undercover instruments
The helper-functions test was per-test reloading system-defaults.el
via `(load ...)' inside a `cl-letf' sandbox that stubs the
side-effecting primitives (server-start, set-locale-environment,
etc). Tests passed, but the coverage gauge stayed stuck at 1/12
because undercover.el only instruments the first load of a matching
source; subsequent re-loads inside test bodies don't get tracked,
so the function bodies showed as uncovered even though every test
called them.
Rewrite the test to call `(require 'system-defaults)' once at top
level, wrapped in the same `cl-letf' stubs. The functions get
instrumented exactly once. Drop the now-unused per-test sandbox
macro. Add two more tests for the `(when (memq ...))' list-without-
comp guard and the non-string-message format branch so coverage
reaches 12/12.
(`test-system-defaults-vc-follow-symlinks.el' still uses the
per-test `(load ...)' pattern because that test *is* the
load-side-effect verification, not a function-body test.)
Diffstat (limited to 'tests/test-setup-telega.bats')
0 files changed, 0 insertions, 0 deletions
