T7.10 - D5: one analytics implementation, and its report on the admin console
Usage analytics existed as five of the nine colliding globals creator-frame.md counted (ANALYTICS_KEY, analyticsLoad, analyticsSave, downloadAnalytics, showAnalytics), twice - and the wizard's copy had no caller, because the button lived on the creator. The admin console had a THIRD private reader (usageLoad/downloadUsage) that only saw the wizard's key. Now: ONE core, html/wp-usage.js (window.WPUsage: load/save/track/download + the two pre-move storage keys, verbatim). The creator and wizard keep only a thin track() wrapper - page state like the creator's dev-mode pause belongs to the page - and record exactly what they recorded before, under the same keys, so everything captured before this task still reads (probe plants a legacy-format event and finds it in the report). The "Usage data" button left the creator toolbar; the report lives in admin.html's usage card, covering BOTH tools with a download each, behind the same admin gate as the rest of the console (a non-admin sees the denied card and nothing else), usable at 390px. Two probes re-pointed, both with the reason in the code: - cards_check pinned admin.js byte-identical to HEAD - right for T6.5, but as a standing probe it would fail every legitimate later edit; D5 targets admin.js by name. A7's localization is protected by the feature checks and the end-to-end drive, plus a wiring assertion on the block itself. - frame_check listed "Usage data" among the toolbar buttons that must be visible; it now asserts the button is GONE, so the duplicate cannot quietly return. Verification (each probe run alone): NEW tests/usage_check.py 15/15 (grep half: WPUsage defined once, no page touches the keys directly, none of the five globals survives anywhere). Regressions: cards_check ALL PASS, frame_check 39/39. Items: D5 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
@@ -235,10 +235,17 @@ def run(page, base, tok, db_path):
|
||||
for needle in ("Localization defaults", "L10N_LOCALES", "L10N_ZONES",
|
||||
"fillLocalization", "saveLocalization", "set-locale", "set-tz"):
|
||||
chk("admin.js still has %s" % needle, needle in src)
|
||||
changed = subprocess.run(
|
||||
["git", "diff", "--stat", "HEAD", "--", "html/admin.js"],
|
||||
cwd=ROOT, capture_output=True, text=True).stdout.strip()
|
||||
chk("...and this task changed nothing in it at all", not changed, changed)
|
||||
# Re-pointed at T7.10, not relaxed. This asserted admin.js was byte-identical
|
||||
# to HEAD - right for T6.5, whose task touched nothing there, but as a
|
||||
# standing probe it failed every LEGITIMATE later edit (D5 moved the usage
|
||||
# report into admin.js by name). A7's protection is the feature checks above
|
||||
# plus the end-to-end localization drive - so pin the localization BLOCK
|
||||
# instead: its functions must not merely exist, they must be uncalled by
|
||||
# nothing, i.e. still wired to the controls that ship the feature.
|
||||
chk("...and the localization block is still wired to its controls",
|
||||
"fillLocalization()" in src
|
||||
and 'onclick="saveLocalization()"' in src
|
||||
and "set-locale" in src)
|
||||
|
||||
|
||||
def main():
|
||||
|
||||
Reference in New Issue
Block a user