test(ip): expand is_ipv6 coverage across the IPv6address alternatives

[vtushar06]

I added 82 tests for sourcemeta::core::is_ipv6, organised around the nine IPv6address alternatives in RFC 3986 Section 3.2.2. Core already passes all 82, so this is coverage, not a fix.

Changes

Everything goes into the existing test/ip/ipv6_test.cc, after the current checks - no new file.

What is new

The grammar admits exactly 59 distinct shapes, counting each (alternative, groups written before the elision, ls32 form) triple once. The file covers a valid case for 29 of them. The other 30 are the backbone of this change, named the way the file already names things - how many groups sit either side of the elision, and whether the low 32 bits are written as an embedded IPv4:

• the third through seventh alternatives at every legal prefix length, in both ls32 forms (::1:2:3:4:5:6, 1::2:3:4:5:6, 1:2::3:4:5:6, ::1:2:3:1.2.3.4, 1:2::3:4:5:1.2.3.4, 1:2:3::4:1.2.3.4, and the rest)
• the eighth and ninth alternatives at the prefix lengths the file skips (1:2:3::4, 1:2:3:4:5::6, 1:2::, 1:2:3::, 1:2:3:4::, 1:2:3:4:5::, 1:2:3:4:5:6::)

On top of the shapes:

• the five dec-octet alternatives exercised in the embedded position rather than through is_ipv4 alone (0, 42, 199, 249, 250, 255), since the interaction is what is untested here - plus an out-of-range and a leading-zero octet at each of the four positions
• hex-digit width and case inside a compressed address (::a, ::ab, ::abc, ::abcd, ::0a, ::00a, ::000a, ::0000, 2001:DB8::AB, 2001:dB8::aB); the file's existing width and case checks are all on the uncompressed eight-group form
• the boundary neighbours of those shapes: eight groups alongside a :: at either end, a dotted quad before an elision (1.2.3.4::) or before another group (::1.2.3.4:5), five hex digits in the first and in a post-elision group, an empty interior group, a quadruple colon, and the three-, five- and trailing-dot embedded IPv4 forms
• the delimiters a delegating parser tends to accept: a half bracket at either end, an RFC 6874 percent-encoded zone identifier (fe80::1%25eth0) and an empty one, a prefix length, and the terminator set (\n, \r, \t, leading and embedded space)
• non-ASCII digit look-alikes a folding parser accepts: fullwidth Ａ, fullwidth digits, an Arabic-Indic digit, a Bengali digit inside the embedded IPv4, and a fullwidth colon

How I checked the expected values

I transcribed the grammar into four independent oracles and required all four to agree before any case was written out: a regex over the RFC 3986 ABNF, a generic ABNF interpreter driven by the rules as data, one derived from the RFC 4291 Section 2.2 prose rather than the grammar, and a bit-level parser that builds the 128-bit value and accepts only if the parse consumes the whole input. They agree on 19,319 structured inputs and on 2,000,000 mutations, with no split. Where the bit-level parser and Python's ipaddress both accept, they agree on the value and not merely the verdict.

I then ran all 82 inputs through the real is_ipv6 - a small driver linking the current ipv6.cc and ipv4.cc, reading hex-encoded bytes so control characters and non-ASCII survive the transport - and got 0 mismatch. The generator refuses to emit any case Core disagrees with, so the block cannot contain a failing test.

The full ip unit suite passes: 424 tests, 0 failures.

Rebased on main after #2713, so the two cases it added that overlap with mine (:1::2 and 1::2:) are not duplicated here.

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[augmentcode]

🤖 Augment PR Summary

Summary: This PR expands unit coverage for sourcemeta::core::is_ipv6 without changing production behavior.

Changes:

• Adds 82 cases to the existing test/ip/ipv6_test.cc suite.
• Covers the remaining RFC 3986 IPv6address compressed-form alternatives.
• Exercises embedded IPv4 tails across decimal-octet boundaries and positions.
• Validates compressed hex widths, leading zeroes, and mixed/uppercase hexadecimal.
• Adds malformed group, compression, embedded-IPv4, bracket, zone-ID, and prefix cases.
• Checks rejection of whitespace, control characters, and non-ASCII look-alikes.

Technical Notes: The tests target the documented strict address-only API: brackets and URI-specific extensions are intentionally rejected; no source implementation changes are included.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. No suggestions at this time.

Comment augment review to trigger a new review at any time.

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}

[jviotti]

Thanks @vtushar06 ! Notice there is a new CLA requirement to submit once through the link in the comment above ^. Once you do it, you are cleared in all Sourcemeta projects.