[Bug]: URL allowlist patterns fail to match deep links due to hard end-anchor in urlPatternToRegex

[MWG-Logan]

Summary

urlPatternToRegex() (duplicated in scripts/content.js and options/options.js) appends a hard $ end-anchor to any non-wildcard pattern. This makes an allowlisted host match only the bare root URL, never a real navigated deep link. An admin who allowlists the exact host of a page can still see that page scanned/blocked, because the actual URL includes a path.

Affected code

// Add end anchor if pattern doesn't end with wildcard
if (!pattern.endsWith("*") && !escaped.endsWith(".*")) {
  escaped = escaped + "$";
}

Worked example

Allowlist entry: https://client.my.salesforce-setup.com/

Produced regex: ^https://client\.my\.salesforce-setup\.com/$

Tested URL (real deep link): https://client.my.salesforce-setup.com/lightning/setup/...

Result: no match. The trailing $ sits immediately after the /, so only the bare root URL matches. The admin allowlisted the exact host, including the protocol-qualified form, yet checkUserUrlAllowlist() returns false and the page-level early-exit never fires.

Scope of this fix (agreed)

Minimal: relax only the trailing anchoring so a host or root URL pattern (with or without a trailing slash) also matches an optional trailing path, query, or fragment. This does not add leading protocol or subdomain tolerance, so bare-domain entries such as five9.com still require the documented https://.../ * form. That broader behavior is intentionally out of scope here.

Expected behavior

• https://host/ and https://host match https://host, https://host/, and https://host/any/deep/link.
• Suffix and prefix tricks must not match. For example https://host.evil.com/ must not be matched by an allow entry for https://host/.
• Existing wildcard patterns (https://google.com/*) and raw regex patterns (^https://login\.microsoftonline\.com$) are unchanged.

Notes

The two copies of urlPatternToRegex in content.js and options.js must stay in sync.