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 b5c266b
Browse filesBrowse the repository at this point in the historyBrowse files
fix(blocklist): make block, allow and unblock see every spelling
After 680ee83 screening matched a national-format number's +1 form,
but the edit paths still worked on one spelling. A refuting review
reproduced three regressions: an older allow of +1 649 555 0123 beat a
newer block of 649 555 0123 (it was blocked before 680ee83), Remove my
rule on the national form left a +1 block in force while saying it was
removed, and a temporary allow of the national form was accepted
against a permanent +1 block that kept winning.
PhoneIdentityCanonicalizer.equivalentForms() now lists every spelling a
line can be stored under: the canonical one, its +1 form, and the
national spellings canonicalization leaves alone (only numbers
libphonenumber rejects, so a valid number still has one form and costs
no extra queries). Screening (CheckContext.alternateForms replaces
e164Fallback) and the outgoing risk warning match all of them, and it
now works from the +1 side too. blockNumber clears allows, addToWhitelist
resolves blocks and refuses a temporary allow against a permanent block,
and unblockByNumber removes blocks, in every form. Lookup's Remove my
rule and Number Detail's unblock go through unblockByNumber, and Number
Detail's blocked state matches every form.
Tests: PhoneIdentityCanonicalizerTest adds the pure equivalentForms
cases. NationalFormatLookupTest adds a case for each regression plus the
reverse spelling, and its precondition now expects the 11-digit
national spelling among the forms (the helper's list grew). Single-form
edits failed four of the new cases; single-form screening failed eight.
0 commit comments