You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 6a532e6
Browse filesBrowse the repository at this point in the historyBrowse files
test(compare, guard, a11y, snapshots): the review's notes on the CI fixtures: paths read as spelled, drive roots, the Linux Trash answer in any order, every view's errors caught
An independent review of 981add4, 28e6098 and b788ae9 found nothing at high or medium. Its
notes:
- compareUnlisted read each path through path.relative, which reads a doubled separator away,
so a store that spelled "//Mail" passed. It now reads the path exactly as spelled, and the
fixtures also root a whole drive's scan ("/" and "C:\") and a share's ("\\srv\share\"), whose
roots already end in their separator: with the store's join for such a root broken, the
whole-drive spelling now fails. The comment says four tests failed on Windows, not every one.
- realMachineGuard named the home Trash folder only as the first failure. On a Linux machine with
another drive whose Trash folder cannot be looked at, that drive is named first. The test now
asks that the home folder is among the failures, and still that gio is not.
- hiddenBoxesSaidAloud took "the registry caught nothing" as proof the Treemap view's unmount ran
to its end. It now counts the unmount's last step (the mode pills' rings turned off) once per
visit away; with the unmount made to return early, the test fails. And the check that the
registry caught nothing moves from that one test into the harness every view test here shares
(viewsPage): each switch of view and each frame fails, naming the error, if the registry logged
one from a view's mount, unmount or onScanChange. No other view had one; with a name nobody
gave put into Disk City's unmount or the Dashboard's mount, their tests now fail.
- snapshotDiffCutoff rooted its fixture at path.join(path.sep, 'snap-root'), drive-less on
Windows, the spelling that failed compareUnlisted there. It is rooted on its drive now; on a Mac
and on Linux the path is the same as before.
Checked on default Node, Node 20 with a 2 GB heap and pt-BR; the Linux branch of the Trash test
with the child told it runs on Linux, red when gio's entry is added back or the home folder's is
dropped; each mutant restored by SHA-256, the mutants of the built page among them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
assert.deepEqual(caught,[],`${what}: control, the Treemap view's own unmount ran to its end — the registry caught nothing from it`);
2599
+
assert.equal(pillsOff,before+1,`${what}: control, the Treemap view's own unmount ran to its last step, once`);
2577
2600
};
2578
2601
constempty='This folder is empty.';
2579
2602
constcap='The map holds at most 250,000 items, and this folder reached that: 1.2 GB has no rectangle here — drill in, or open the Grid view, to see the rest.';
assert.equal(result.failed[0]?.location,homeTrash,`and the answer names the Trash folder the fallback could not clear: ${JSON.stringify(value)}`);
637
+
assert.ok(result.failed.some((f)=>f.location===homeTrash),`and the answer names the Trash folder the fallback could not clear: ${JSON.stringify(value)}`);
637
638
assert.ok(!result.failed.some((f)=>f.location===cmd),`and not gio, which the fallback stood in for: ${JSON.stringify(value)}`);
0 commit comments