[maximilianfv]

What

On macOS, getaddrinfo with ai_family = AF_INET6 and no flags answers an IPv4 literal such as "127.0.0.1" with the IPv4-mapped address ::ffff:127.0.0.1. lookup_host passes that on, so sceNetResolverStartNtoa6(resolver, "127.0.0.1", ...) returns 0 and writes ::ffff:127.0.0.1. guest_sce_net expects SCE_NET_ERROR_RESOLVER_ENODNS (0x804101E1) with the output buffer unchanged (core/libs/tests/GuestSceNet.cpp:1932), and the sceNetResolverStartNtoa6 entry in docs/dev/debt/libSceNet.md says a name with no IPv6 address fails with RESOLVER_ENODNS.

FreeBSD 11.4's getaddrinfo returns no IPv4-mapped address for an AF_INET6 query unless AI_V4MAPPED is set: an IPv4 literal is accepted only with that flag (explore_numeric), and DNS asks for A records only with that flag (_dns_getaddrinfo). lookup_host never sets it.

This PR makes lookup_host skip IPv4-mapped results of an AF_INET6 lookup on macOS (IN6_IS_ADDR_V4MAPPED). sceNetResolverStartNtoa6 is the only AF_INET6 caller of lookup_host. Linux and Windows are unchanged.

A small host probe (getaddrinfo, AF_INET6, flags 0) on macOS returned ::ffff:127.0.0.1 for "127.0.0.1", ::1 for "::1" and ::1 for "localhost".

Limitations, on macOS only:

• An AAAA record whose value is itself an IPv4-mapped address is now skipped. FreeBSD 11.4 returns it: the filter for such records is behind FILTER_V4MAPPED, which its libc Makefiles do not define.
• A name with only A records, looked up as AF_INET6, now fails with RESOLVER_ENODNS as on FreeBSD (only an IPv4 literal was probed on macOS, not a real A-only name).

Prior art: the same change was written before in two whole-port PRs that are now closed, so both authors are credited as co-authors on the commit. #7332 by Ycaro Gabriel (ycarogabriel-lgtm) has identical code in commit a8fa3e9 (2026-10-17). #8658 by Himansh Raj (iamthehimansh) has identical code plus a comment in commit 301073a (2026-10-19). Open #7201 by marion909 (commit c31e82f, 2026-10-16) takes a different approach for the same test: on macOS, an AF_INET6 lookup of a name that parses as an IPv4 literal fails before getaddrinfo runs. That approach does not cover A-only names. Apart from the whole-port #7201, no open PR changes lookup_host.

Tested

Apple M1 Pro, macOS 27.0.1, x86_64 build under Rosetta 2.

main alone does not build on macOS, so I tested in a tree made of mugurc's macOS series (ci/macos-series-2, 27e968974) merged with main at 1e20650e84, open #9231 (7c5d97324) and my own sceNetIoctl FIONREAD fix (not yet opened). Only this change was toggled between runs.

• ctest -R '^(guest_sce_net|guest_resolver|guest_net_ctl|guest_inet_address|guest_net_.*|guest_datagram_.*|net_dontwait_.*|guest_socket_error|guest_udp)$' (22 tests): 21 pass with and without the change.
• guest_sce_net aborts with and without the change. Without the change it stops at the check above (GuestSceNet.cpp:1932 on current main, located with lldb on an -O0 rebuild). With the change it passes that check and stops later, at the sceNetCtlGetIfStat(&first) call (GuestSceNet.cpp:2168 on current main). That stop is an uncaught "sceNetCtlGetIfStat: host interface byte counters are not implemented on this platform" (libSceNetCtl/Export.cpp:511), a separate macOS gap that open feat(libSceNetCtl): read macOS interface byte counters #9728 (MarcZX10) closes.

Not tested on Linux or Windows. Their code is unchanged.

Checklist

• Based on current main; no other open PR implements the same functions (except Macos port #7201, the whole macOS port, see What)
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: fix(libSceNet): print the interface and socket tables from sceNetShowIfconfig and sceNetShowNetstat #10502, fix(libSceNet): reject the flags and names the module rejects #9727 and fix(libSceNet): keep MSG_DONTWAIT and aborted stream sends short on macOS hosts #9733 (GuestSceNet.cpp and neighbouring docs/dev/debt/libSceNet.md entries; I rebase this one after them); merge order with Macos port #7201 (lookup_host) asked at Macos port #7201 (comment)
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

🤖 Generated with Claude Code

[MarcZX10]

The prior-art section says #7201 remains open and makes a whole-port change that overlaps the same lookup_host behavior, but this PR's checklist says Depends on: none. CONTRIBUTING.md says the PR overlap comment identifies open PRs that change the same lines or exports; the authors should agree which goes first and declare it with Depends on. Please coordinate with #7201's author and record the agreed order (or explain why this exact overlap does not require one) before merging. I have not independently run the macOS behavior probe; the stated host probe and targeted regression evidence are useful.

[MarcZX10]

The PR body identifies overlapping lookup_host behavior in open #7201 but the checklist still says Depends on: none. CONTRIBUTING requires the authors to agree on merge order and record the dependency. I left the specific request in this comment; no author response or body update is present at this head. I have not approved pending that coordination.

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
PR 11155	docs/dev/debt/libSceNet.md L63-64

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 5836	core/libs/prx/libSceNet/Export.cpp docs/dev/debt/libSceNet.md
PR 7201	core/libs/prx/libSceNet/Export.cpp
PR 9932	core/libs/prx/libSceNet/Export.cpp docs/dev/debt/libSceNet.md
PR 9948	core/libs/prx/libSceNet/Export.cpp docs/dev/debt/libSceNet.md
PR 11341	core/libs/prx/libSceNet/Export.cpp

[MarcZX10]

Follow-up on the current overlap warnings at head b5868dd32860621102859b462b6cc0486794994a. Manual checks confirm two coordination items beyond the original checklist mismatch:

• #7201 edits the same lookup_host path on macOS (it rejects numeric IPv4 literals before getaddrinfo; this PR filters IPv4-mapped results). git merge-tree --write-tree upstream/pr/7201 upstream/pr/10629 conflicts inside lookup_host, so these behaviors need an agreed order and a combined test/implementation plan. The body mentions #7201 but the checklist still says Depends on: none.
• #10502 implements sceNetShowIfconfig and sceNetShowNetstat, while this branch's CheckNetstat still expects those calls to throw. The three-way merge conflicts in GuestSceNet.cpp at that test and in the Silent stubs debt entry. Please coordinate order with #10502 and reconcile the test expectations so the combined branch tests the implemented diagnostics rather than treating them as unsupported.

I also confirmed adjacent debt-file conflicts with #9727 and #9733 via merge-tree; their implementation changes are in other functions, but the shared documentation hunks will need to retain both updates. The broad #5836 diff changes peer socket types/options, not lookup_host, so I found no direct resolver implementation overlap there.

Please agree on the merge order with the affected authors and record the required dependency or dependencies in the checklist. I have not run a local macOS resolver test; the PR's host probe and reported platform test remain author evidence.

[maximilianfv]

Thanks. The checklist now has Depends on: #10502, #9727 and #9733: this PR waits for them and I rebase it afterwards, keeping their entries and #10502's implemented CheckNetstat expectations. For lookup_host I asked marion909 about the order with #7201 (link in the checklist).

AI-assisted: yes

[Gonic0]

On the limitation you list about A-only names: I probed it on an M5 Pro, macOS 27.2 (26B5086k), built both x86_64 (Rosetta 2) and arm64. Both builds give the same results. The network is dual-stack with global IPv6 and no DNS64. getaddrinfo with AF_INET6:

Name	flags 0	AI_V4MAPPED
127.0.0.1	::ffff:127.0.0.1	same
::1	::1	same
localhost	::1	same
ipv4only.arpa (A only, RFC 7050)	::ffff:192.0.0.170, ::ffff:192.0.0.171	same
github.com (A only)	::ffff:140.82.114.3	same

So a real A-only name is mapped on macOS too, even with flags 0, and with this change it fails with RESOLVER_ENODNS as on FreeBSD. AI_V4MAPPED makes no difference here, so macOS can't be steered out of mapping by flags; filtering the results is the only way.

libSceNet/Export.cpp at b5868dd328 also compiles here (Apple clang 21.0.0, x86_64 Release), with no warnings in the changed lines under -Wall -Wextra. I didn't run guest_sce_net.

AI-assisted: yes