[Bug]: Antigravity installation hides IPv6-related startup failure on WSL2

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/server

Steps to reproduce

1. Use WSL2 with IPv6 disabled through this setting in %UserProfile%\.wslconfig:

   [wsl2]
   kernelCommandLine=ipv6.disable=1
   

2. Confirm the running kernel has ipv6.disable=1 in /proc/cmdline.
3. With no existing Antigravity installation or executable override, start T3:

   npx t3@0.0.39-nightly.20260904.1280
   

4. Install the Antigravity runtime through T3's provider settings.

Expected behavior

If Antigravity cannot start because IPv6 socket support is disabled, T3 should display that cause and actionable guidance.

The runtime already emits a diagnostic identifying the failed IPv6 check. T3 should preserve that information in its installation error or diagnostic logs.

Actual behavior

Downloading, archive size/SHA-256 verification, and extraction succeed. During startup validation, agy_acp_server.par aborts with SIGABRT after approximately three seconds.

T3 displays only:

The downloaded Antigravity runtime could not start in this environment.

The trace logs expose SIGABRT, but the runtime's explanatory stderr is discarded. Retrying installation repeats the download and failure without explaining how to resolve it.

Running the exact checksum-verified package directly reveals that Antigravity's native initializer requires IPv6 socket support. Creating an AF_INET6 socket in this environment fails with EAFNOSUPPORT.

Impact

Major degradation or frequent failure

Version or commit

t3@0.0.39-nightly.20260904.1, agy_acp_server_1.1.1

Environment

Ubuntu 22.04.5 LTS under WSL2, x86-64., Kernel 6.6.114.1-microsoft-standard-WSL2

Logs or stack traces

T3 trace:


AntigravityInstallationError:
The downloaded Antigravity runtime could not start in this environment.

[cause]: AcpTransportError:
ACP transport operation read-process-exit-status failed.

[cause]: Error:
Process interrupted due to receipt of signal: 'SIGABRT'


Direct runtime stderr:


enforce_kernel_ipv6_support.cc:70] Check failed:
AddressFamilySupported(AF_INET6, &loopback6_ok)
Failed to create an AF_INET6 socket, which likely means that you're
running a kernel with CONFIG_IPV6=n. This configuration is obsolete.
You may temporarily set --enforce_kernel_ipv6_support=false to bypass
this check, but code elsewhere in google3 may fail as result.


The kernel supports IPv6, but it is disabled by the boot parameter.

Screenshots, recordings, or supporting files

No response

Workaround

Remove ipv6.disable=1 from .wslconfig and restart WSL.

Alternatively, launch the runtime through a wrapper that adds:

--enforce_kernel_ipv6_support=false

This bypass was tested with the exact pinned runtime. The ACP initialization response passed every managed-installation validation check, including agent identity, version, protocol, load/resume/logout capabilities, and oauth-personal authentication support. Sign-in also works fine.

For automatic PATH discovery, expose the wrapper as agy_acp_server.par. T3 resolves symlinks and requires localharness_external beside the resolved wrapper.

Start in apps/server and trace the Antigravity installation startup-validation path, then reproduce the failure on WSL2 with IPv6 disabled. Check where the runtime's stderr is discarded and verify that the installation error or diagnostic logs retain the IPv6 failure and available guidance instead of only reporting SIGABRT.