InetAddresses.forString(String) parses non-ASCII decimals

[Marcono1234]

The method InetAddresses.forString(String) parses non-ASCII decimals due to its usage of Character.digit(...).
For example:

// /127.0.0.1
System.out.println(InetAddresses.forString("127.0.0.\u0661"));
// /0:0:0:0:0:0:0:1
System.out.println(InetAddresses.forString("0:0:0:0:0:0:0:\u0661"));

While it appears there is no RFC which strictly specifies the syntax, related RFCs, such as RFC 3986 specifying the URI syntax, explicitly only allow ASCII 0-9 (respectively f for IPv6).
Additionally there is this draft which also only permits ASCII digits.

Note however, that Java's InetAddress.getByName(String) does support non-ASCII digits as well, though it is not clear whether that is intended or just an oversight.

(This issue might be relevant for #3417)

[eamonnmcmanus]

I made the appropriate change and tried it out against the full range of Google's internal tests. I found that some Google projects had tests explicitly verifying that addresses like １２７.０.０.１ work, where those are "fullwidth" digits. We can fix the code in question so that it sanitizes non-ASCII digits before calling InetAddresses.forString (or the various other InetAddresses methods), but it seems a bit risky. I think it would be better to document that these methods accept non-ASCII digits, and explain how to reject them if you don't want that (for example !CharMatcher.ascii().matchesAllOf(address). The fact that java.net.InetAddress and java.net.URL also accept non-ASCII digits does argue in favour of InetAddresses continuing to accept them.

I didn't see any test failures from requiring that the port number in HostAndPort be ASCII, and I think that change might be safer to make. (#5681)