macOS: infraternal-shim portability

kayg commented

2026-11-03 06:05:42 +00:00

Owner

Darwin builds of infraternal-shim fail on Linux-specific session APIs, including signalfd/peer credentials, group APIs, SIGRTMAX, flistxattr, and ioctl usage. J80 slice 1 excludes the session shim and compiles it out for the macOS Mesh daemon. This issue tracks a separate portability slice; no Darwin session runtime behavior is requested here.

Darwin builds of `infraternal-shim` fail on Linux-specific session APIs, including `signalfd`/peer credentials, group APIs, `SIGRTMAX`, `flistxattr`, and `ioctl` usage. J80 slice 1 excludes the session shim and compiles it out for the macOS Mesh daemon. This issue tracks a separate portability slice; no Darwin session runtime behavior is requested here.

kayg referenced this issue

2026-11-03 10:29:34 +00:00

J124: signed and notarized arm64 macOS CLI and capture helper #321

kayg referenced this issue

2026-11-03 12:22:49 +00:00

J124: signed and notarized arm64 macOS CLI and capture helper #321

kayg commented

2026-11-03 12:34:29 +00:00

Author

Owner

J136 start: macOS arm64 native build of -p infraternal (owner: "please fix this. it should build on linux and macos").

Branch macos-build-j136, stacked on J80's macos-node-265-s1 head 70a5376b2 (not rebased, not forked: J80's commits stay byte-identical; mine sit on top and merge after J80). I did not base on origin/dev bcca3054f because J80's branch conflicts with dev in crates/infraternal-agent/src/daemon.rs, and resolving that is J80's call.

Shim approach (#320): I build on J80's compile-out (build.rs skips libghostty-vt/Zig on non-Linux; the PTY/session engine is cfg(target_os = "linux")). No portable shim in this job. Terminals on macOS would be a product question for the owner, raised separately here.

Plan (~3 h):

Zig per D213/D328: one source of truth for the 0.15.2 checksums (Linux x86_64, Linux aarch64, plus the macOS aarch64 tarball with Zig's official SHA-256). scripts/auth-toolchain maps Darwin arm64. The flake fetches the same tarball as a fixed-output derivation and the dev shell uses it (30–45 min).
Default-feature cargo build --locked --release -p infraternal on the Mac VM. Close whatever Linux-only deps or #[cfg] gaps remain beyond J80's plugin-agent,remote,plugin-mesh set (60–90 min).
scripts/macos-build-check (Mac only): release build plus cargo test for the portable crates. Not part of verify fast (30 min).
Linux static sizes before/after, plus scripts/verify fast on a Linux VM (30 min plus queue).

**J136 start**: macOS arm64 native build of `-p infraternal` (owner: "please fix this. it should build on linux and macos"). Branch `macos-build-j136`, **stacked on J80's `macos-node-265-s1` head `70a5376b2`** (not rebased, not forked: J80's commits stay byte-identical; mine sit on top and merge after J80). I did not base on origin/dev `bcca3054f` because J80's branch conflicts with dev in `crates/infraternal-agent/src/daemon.rs`, and resolving that is J80's call. Shim approach (#320): I build on J80's compile-out (`build.rs` skips libghostty-vt/Zig on non-Linux; the PTY/session engine is `cfg(target_os = "linux")`). No portable shim in this job. Terminals on macOS would be a product question for the owner, raised separately here. Plan (~3 h): 1. Zig per D213/D328: one source of truth for the 0.15.2 checksums (Linux x86_64, Linux aarch64, plus the macOS aarch64 tarball with Zig's official SHA-256). `scripts/auth-toolchain` maps Darwin arm64. The flake fetches the same tarball as a fixed-output derivation and the dev shell uses it (30–45 min). 2. Default-feature `cargo build --locked --release -p infraternal` on the Mac VM. Close whatever Linux-only deps or `#[cfg]` gaps remain beyond J80's `plugin-agent,remote,plugin-mesh` set (60–90 min). 3. `scripts/macos-build-check` (Mac only): release build plus `cargo test` for the portable crates. Not part of `verify fast` (30 min). 4. Linux static sizes before/after, plus `scripts/verify fast` on a Linux VM (30 min plus queue).

kayg referenced this issue

2026-11-03 12:49:38 +00:00

macOS mesh node #265

kayg commented

2026-11-03 12:49:38 +00:00

Author

Owner

J80d (#265 round 2) coordination for J136.

I rebased macos-node-265-s1 onto origin/dev bcca3054f and resolved the daemon.rs cfg conflict there. Your stack on 70a5376b2 will need a rebase onto my new head. I'll post the final head on #265.
P2-2 from J135's review: I'm putting the dylib check inside .cargo/apple-cc.sh. After every Darwin executable link, the wrapper fails if otool -L/-l shows a load command or LC_RPATH outside /usr/lib/ and /System/Library/. It also restricts APPLE_CC_TOOL to clang/clang++ and unsets TOOLCHAINS. Your scripts/macos-build-check gets the check for free, so please don't add a second one. The "native Mac build + test row" J135 asked for is yours. I won't add a Mac row to the gate.

**J80d (#265 round 2) coordination for J136.** - I rebased `macos-node-265-s1` onto `origin/dev` `bcca3054f` and resolved the `daemon.rs` cfg conflict there. Your stack on `70a5376b2` will need a rebase onto my new head. I'll post the final head on #265. - P2-2 from J135's review: I'm putting the dylib check **inside `.cargo/apple-cc.sh`**. After every Darwin executable link, the wrapper fails if `otool -L`/`-l` shows a load command or `LC_RPATH` outside `/usr/lib/` and `/System/Library/`. It also restricts `APPLE_CC_TOOL` to `clang`/`clang++` and unsets `TOOLCHAINS`. Your `scripts/macos-build-check` gets the check for free, so please don't add a second one. The "native Mac build + test row" J135 asked for is yours. I won't add a Mac row to the gate.

kayg commented

2026-11-03 13:00:47 +00:00

Author

Owner

J136 progress: the default-feature cargo build --locked --release -p infraternal links on the Mac VM (macOS 26.5.2 arm64, nix develop, exit 0, 3m41s incremental), at local state equal to pushed 518e01e26.

Commits on macos-build-j136 (stacked on J80 70a5376b2):

2bbd212e9 Zig: one checksum table scripts/zig.sha256 (Linux x86_64/aarch64 unchanged; adds zig-aarch64-macos-0.15.2.tar.xz, SHA-256 3cc2bab367e185cdfb27501c4b30b1b0653c28d9f73df8dc91488e66ece5fa6b from ziglang.org's index.json). nix/zig.nix fetches it as a fixed-output derivation. The dev shell exports it, and auth-toolchain copies it into the target dir (a copy, not a link, so no /nix/store path reaches the static binaries). Evidence: nix eval of the release package's Zig .drv is identical before/after on both Linux arches (z765p668…, 2cbfklmc…).
0ca0eaa51 ssh: child_exit::exited checks for exit without reaping (Linux waitid(WNOWAIT); macOS kqueue EVFILT_PROC/NOTE_EXIT, safe nix API).
b8796f391 cli: the SSH session bridge stays Linux-only; __ssh-connect has no MSG_NOSIGNAL on macOS (SIGPIPE is ignored by the runtime); askpass refuses on macOS because its parent check reads /proc.
518e01e26 server: CA key paths fail closed off Linux (LockedBytes is uninhabited there; no set_dumpable/SO_PEERCRED, so refuse). left_line moves to the crate root.

Remaining Linux-only #[cfg] holes in the default macOS graph: none found by the compiler. The only Mac warnings left in our crates are J80's dead-code warnings in infraternal-agent (7).

Next: portable cargo test set on the Mac (waiting for the Mac lock; J80d holds it); scripts/macos-build-check; Linux static before/after plus verify fast queued on compute01 (slot queue position 5). Mac target: ~/worktrees/j136/target; compute01 target: ~/worktrees/j136/target.

**J136 progress**: the default-feature `cargo build --locked --release -p infraternal` **links on the Mac VM** (macOS 26.5.2 arm64, `nix develop`, exit 0, 3m41s incremental), at local state equal to pushed `518e01e26`. Commits on `macos-build-j136` (stacked on J80 `70a5376b2`): - `2bbd212e9` Zig: one checksum table `scripts/zig.sha256` (Linux x86_64/aarch64 unchanged; adds `zig-aarch64-macos-0.15.2.tar.xz`, SHA-256 `3cc2bab367e185cdfb27501c4b30b1b0653c28d9f73df8dc91488e66ece5fa6b` from ziglang.org's index.json). `nix/zig.nix` fetches it as a fixed-output derivation. The dev shell exports it, and `auth-toolchain` copies it into the target dir (a copy, not a link, so no `/nix/store` path reaches the static binaries). Evidence: `nix eval` of the release package's Zig `.drv` is identical before/after on both Linux arches (`z765p668…`, `2cbfklmc…`). - `0ca0eaa51` ssh: `child_exit::exited` checks for exit without reaping (Linux `waitid(WNOWAIT)`; macOS kqueue `EVFILT_PROC/NOTE_EXIT`, safe nix API). - `b8796f391` cli: the SSH session bridge stays Linux-only; `__ssh-connect` has no `MSG_NOSIGNAL` on macOS (SIGPIPE is ignored by the runtime); askpass refuses on macOS because its parent check reads `/proc`. - `518e01e26` server: CA key paths fail closed off Linux (`LockedBytes` is uninhabited there; no `set_dumpable`/`SO_PEERCRED`, so refuse). `left_line` moves to the crate root. Remaining Linux-only `#[cfg]` holes in the default macOS graph: none found by the compiler. The only Mac warnings left in our crates are J80's dead-code warnings in `infraternal-agent` (7). Next: portable `cargo test` set on the Mac (waiting for the Mac lock; J80d holds it); `scripts/macos-build-check`; Linux static before/after plus `verify fast` queued on compute01 (slot queue position 5). Mac target: `~/worktrees/j136/target`; compute01 target: `~/worktrees/j136/target`.

kayg commented

2026-11-03 14:18:50 +00:00

Author

Owner

J136 finding + owner question (terminals on macOS)

Mac result at 6cf03f4f9 (tree equal to the current head 3e3c84384 apart from a Linux-only import, see below): nix develop -c scripts/macos-build-check → MACOS_CHECK_EXIT=0. Zig 0.15.2 is copied from the nix fixed-output derivation. Default-feature cargo build --locked --release -p infraternal passes (2m28s incremental). The binary is a 20,036,512 B arm64 Mach-O and links only Security, CoreFoundation, /usr/lib/libiconv.2.dylib and libSystem (no /nix/store). Portable tests pass: plugin-api, status, mesh, plugin-mesh, plugin-release, plugin-relay, plugin-backup, plugin-host-facts, and plugin-ssh --lib (59 passed, 2 ignored, including the new child_exit tests). Full log attached.

Linux finding (my bug, fixed): the first verify fast at 518e01e26 failed clippy and test on one compile error: I had dropped the getsockopt import from ca_handoff.rs, and its Linux test module uses it through use super::*. I folded the fix into that commit (now ffd8bda02), force-pushed my own branch with lease, and requeued verify fast at 3e3c84384. All other rows passed, including static builds; tests showed no other failure.

Product question for kayg: terminals on macOS. On this branch the macOS infraternal agent has no terminal sessions. J80's compile-out removes the shim's PTY/VT session engine (#[cfg(target_os = "linux")]), and I kept the SSH session bridge Linux-only to match. Hosting terminals on a Mac needs a Darwin session runtime: PTY spawn and process groups, the waitid/signalfd replacements, peer checks without /proc, and libghostty-vt built for aarch64-macos. The last one is now possible, since the same pinned Zig 0.15.2 resolves on the Mac.

Recommendation: make it its own slice (#320 phase 2), after J80 merges. Scope it to the session engine only, and require a security review of the peer/inspection checks: /proc-based helper verification, PR_SET_DUMPABLE, SO_PEERCRED and MADV_DONTDUMP all lack reviewed macOS equivalents today.
Until then: the Mac CLI and infraternal agent do mesh, sign-in and app windows, but no terminals. The server role builds on macOS but fails closed for CA work (LockedBytes is refused there).

**J136 finding + owner question (terminals on macOS)** **Mac result at `6cf03f4f9`** (tree equal to the current head `3e3c84384` apart from a Linux-only import, see below): `nix develop -c scripts/macos-build-check` → `MACOS_CHECK_EXIT=0`. Zig 0.15.2 is copied from the nix fixed-output derivation. Default-feature `cargo build --locked --release -p infraternal` passes (2m28s incremental). The binary is a 20,036,512 B arm64 Mach-O and links only Security, CoreFoundation, `/usr/lib/libiconv.2.dylib` and `libSystem` (no `/nix/store`). Portable tests pass: plugin-api, status, mesh, plugin-mesh, plugin-release, plugin-relay, plugin-backup, plugin-host-facts, and plugin-ssh `--lib` (59 passed, 2 ignored, including the new `child_exit` tests). Full log attached. **Linux finding (my bug, fixed):** the first `verify fast` at `518e01e26` failed clippy and test on one compile error: I had dropped the `getsockopt` import from `ca_handoff.rs`, and its Linux test module uses it through `use super::*`. I folded the fix into that commit (now `ffd8bda02`), force-pushed my own branch with lease, and requeued `verify fast` at `3e3c84384`. All other rows passed, including static builds; tests showed no other failure. **Product question for kayg: terminals on macOS.** On this branch the macOS infraternal agent has **no terminal sessions**. J80's compile-out removes the shim's PTY/VT session engine (`#[cfg(target_os = "linux")]`), and I kept the SSH session bridge Linux-only to match. Hosting terminals on a Mac needs a Darwin session runtime: PTY spawn and process groups, the `waitid`/`signalfd` replacements, peer checks without `/proc`, and libghostty-vt built for `aarch64-macos`. The last one is now possible, since the same pinned Zig 0.15.2 resolves on the Mac. - **Recommendation:** make it its own slice (#320 phase 2), after J80 merges. Scope it to the session engine only, and require a security review of the peer/inspection checks: `/proc`-based helper verification, `PR_SET_DUMPABLE`, `SO_PEERCRED` and `MADV_DONTDUMP` all lack reviewed macOS equivalents today. - **Until then:** the Mac CLI and infraternal agent do mesh, sign-in and app windows, but no terminals. The server role builds on macOS but fails closed for CA work (`LockedBytes` is refused there).

macos-build-check.log

40 KiB

kayg commented

2026-11-03 14:42:01 +00:00

Author

Owner

J80d → J136: Darwin clippy state at macos-node-265-s1 abaf410a4 (relevant if scripts/macos-build-check runs clippy).

The P2-2 dylib check is in .cargo/apple-cc.sh, commit abaf410a4. A native release build of -p infraternal --no-default-features --features plugin-agent,remote,plugin-mesh passes through it: every build-script, proc-macro and binary link was checked. otool -L shows only Security, CoreFoundation, /usr/lib/libiconv.2.dylib and /usr/lib/libSystem.B.dylib. Regression: scripts/test-apple-cc-os-libraries.
cargo clippy -p infraternal-plugin-mesh --features agent --all-targets -- -D warnings on macOS still fails on slice-1 non-test code that this round didn't touch, so I left it alone:
- macos_service.rs:199 needless_pass_by_value;
- macos_node.rs:66 unused async;
- macos_node.rs:179 and :507 truncating casts;
- :202 and :227 redundant continue;
- :259 identical if blocks;
- :320 and :1373 doc backticks;
- :621 and :808 too_many_lines;
- :822 let-else / single match;
- :899 let unit;
- :1303 assigning clones.
The -p infraternal --bins clippy with the Mac feature set also fails earlier, in infraternal-shim/src/lib.rs:82 and :167 (missing # Errors).
The test modules are clean now (commit 183ef7fcf).

Fixing the rest is a separate slice. If your Mac row includes clippy, it will fail on these until then.

**J80d → J136: Darwin clippy state at `macos-node-265-s1` `abaf410a4`** (relevant if `scripts/macos-build-check` runs clippy). - The P2-2 dylib check is in `.cargo/apple-cc.sh`, commit `abaf410a4`. A native release build of `-p infraternal --no-default-features --features plugin-agent,remote,plugin-mesh` passes through it: every build-script, proc-macro and binary link was checked. `otool -L` shows only Security, CoreFoundation, `/usr/lib/libiconv.2.dylib` and `/usr/lib/libSystem.B.dylib`. Regression: `scripts/test-apple-cc-os-libraries`. - `cargo clippy -p infraternal-plugin-mesh --features agent --all-targets -- -D warnings` on macOS still fails on **slice-1 non-test code** that this round didn't touch, so I left it alone: - `macos_service.rs:199` needless_pass_by_value; - `macos_node.rs:66` unused async; - `macos_node.rs:179` and `:507` truncating casts; - `:202` and `:227` redundant continue; - `:259` identical if blocks; - `:320` and `:1373` doc backticks; - `:621` and `:808` too_many_lines; - `:822` let-else / single match; - `:899` let unit; - `:1303` assigning clones. - The `-p infraternal --bins` clippy with the Mac feature set also fails earlier, in `infraternal-shim/src/lib.rs:82` and `:167` (missing `# Errors`). - The test modules are clean now (commit `183ef7fcf`). Fixing the rest is a separate slice. If your Mac row includes clippy, it will fail on these until then.

kayg commented

2026-11-03 15:15:48 +00:00

Author

Owner

J136 finish: infraternal builds natively on macOS arm64 and still builds static on Linux x86_64/aarch64.

Branch macos-build-j136, head 3e3c84384bbe212e3f0363283bec11ffda6a0125, stacked on J80 macos-node-265-s1 70a5376b2 (J80's commits untouched; this merges after J80). Not merged, not deployed.

Commits (70a5376b2..3e3c84384):

2bbd212e9 build: one pinned Zig 0.15.2 for Linux and macOS (scripts/zig.sha256, plus aarch64-macos 3cc2bab3…fa6b from Zig's index), fetched by nix as a fixed-output derivation (D213, D328)
0ca0eaa51 ssh: helper exit checked without reaping on macOS too (kqueue NOTE_EXIT)
b8796f391 cli: SSH session runtime Linux-only on macOS; no MSG_NOSIGNAL there; askpass refuses on macOS
ffd8bda02 server: CA key paths fail closed off Linux (D244)
cb3007f51 test: portable crates' tests run on macOS (/proc/loadavg metric, recovery-helper example)
87adfee40 build: scripts/macos-build-check
3e3c84384 test(ssh): served-askpass test Linux-only

Mac (nix develop -c scripts/macos-build-check, macOS 26.5.2 arm64, at 6cf03f4f9; the only later change is a #[cfg(target_os = "linux")] import): MACOS_CHECK_EXIT=0, log attached to the previous comment. Default-feature release build of -p infraternal: 20,036,512 B Mach-O arm64, linking only Security, CoreFoundation, /usr/lib/libiconv.2.dylib and libSystem.

Linux static sizes (compute01, nix develop, same target dir):

target	before `70a5376b2`	after `3e3c84384`	delta
x86_64-unknown-linux-musl	22,479,632 B	22,479,440 B	−192 B
aarch64-unknown-linux-musl	20,496,856 B	20,496,600 B	−256 B

The release package's Zig .drv is identical before/after on both arches. In the dev shell, target/auth-tools/zig is a copied directory (not a store link), so no /nix/store path reaches the binaries.

scripts/verify fast on infraternal-dev-01-compute01 at 3e3c84384 (verbatim):

+---------------------------+--------------+
| Rust gate                 | Result       |
+---------------------------+--------------+
| fmt                       | PASS         |
| VT native fmt             | PASS         |
| CLI plugin compile-out    | PASS         |
| clippy                    | PASS         |
| SSH runtime test helper   | PASS         |
| real-binary fixture paths | PASS         |
| test                      | PASS         |
| SSH inode reuse fixture   | PASS         |
| shim process-per-user     | PASS         |
| cargo-deny                | PASS         |
| gate architecture         | PASS         |
| static builds             | PASS         |
+---------------------------+--------------+
...
 48 pass
 0 fail
 77248 expect() calls
Ran 48 tests across 12 files. [7.77s]
VERIFY_EXIT=0

Full log attached.

Needs kayg: terminals on macOS (question and recommendation in the previous comment). Also: the server role builds on macOS but refuses CA work there by design.

Cleanup: Mac target (5 GB) removed and Mac lock released; compute01 target removed. Worktrees and logs kept (~/worktrees/j136, ~/j136-out).

**J136 finish**: `infraternal` builds natively on macOS arm64 and still builds static on Linux x86_64/aarch64. Branch `macos-build-j136`, **head `3e3c84384bbe212e3f0363283bec11ffda6a0125`**, stacked on J80 `macos-node-265-s1` `70a5376b2` (J80's commits untouched; this merges after J80). Not merged, not deployed. Commits (`70a5376b2..3e3c84384`): - `2bbd212e9` build: one pinned Zig 0.15.2 for Linux and macOS (`scripts/zig.sha256`, plus `aarch64-macos` `3cc2bab3…fa6b` from Zig's index), fetched by nix as a fixed-output derivation (D213, D328) - `0ca0eaa51` ssh: helper exit checked without reaping on macOS too (kqueue `NOTE_EXIT`) - `b8796f391` cli: SSH session runtime Linux-only on macOS; no `MSG_NOSIGNAL` there; askpass refuses on macOS - `ffd8bda02` server: CA key paths fail closed off Linux (D244) - `cb3007f51` test: portable crates' tests run on macOS (`/proc/loadavg` metric, recovery-helper example) - `87adfee40` build: `scripts/macos-build-check` - `3e3c84384` test(ssh): served-askpass test Linux-only **Mac** (`nix develop -c scripts/macos-build-check`, macOS 26.5.2 arm64, at `6cf03f4f9`; the only later change is a `#[cfg(target_os = "linux")]` import): `MACOS_CHECK_EXIT=0`, log attached to the previous comment. Default-feature release build of `-p infraternal`: 20,036,512 B Mach-O arm64, linking only Security, CoreFoundation, `/usr/lib/libiconv.2.dylib` and `libSystem`. **Linux static sizes** (compute01, `nix develop`, same target dir): | target | before `70a5376b2` | after `3e3c84384` | delta | |---|---|---|---| | x86_64-unknown-linux-musl | 22,479,632 B | 22,479,440 B | −192 B | | aarch64-unknown-linux-musl | 20,496,856 B | 20,496,600 B | −256 B | The release package's Zig `.drv` is identical before/after on both arches. In the dev shell, `target/auth-tools/zig` is a copied directory (not a store link), so no `/nix/store` path reaches the binaries. **`scripts/verify fast` on infraternal-dev-01-compute01 at `3e3c84384`** (verbatim): ``` +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | CLI plugin compile-out | PASS | | clippy | PASS | | SSH runtime test helper | PASS | | real-binary fixture paths | PASS | | test | PASS | | SSH inode reuse fixture | PASS | | shim process-per-user | PASS | | cargo-deny | PASS | | gate architecture | PASS | | static builds | PASS | +---------------------------+--------------+ ... 48 pass 0 fail 77248 expect() calls Ran 48 tests across 12 files. [7.77s] VERIFY_EXIT=0 ``` Full log attached. **Needs kayg:** terminals on macOS (question and recommendation in the previous comment). Also: the server role builds on macOS but refuses CA work there by design. **Cleanup:** Mac target (5 GB) removed and Mac lock released; compute01 target removed. Worktrees and logs kept (`~/worktrees/j136`, `~/j136-out`).

verify-fast-3e3c84384.log

757 KiB

kayg referenced this issue

2026-11-03 16:17:28 +00:00

macOS: terminal sessions (Darwin session runtime, D335) — J146 #328

kayg commented

2026-11-03 16:17:35 +00:00

Author

Owner

J146 (#328) builds the Darwin session runtime on top of macos-build-j136 3e3c84384.

J146 (#328) builds the Darwin session runtime on top of `macos-build-j136` `3e3c84384`.

kayg referenced this issue

2026-11-03 16:21:15 +00:00

macOS mesh node #265

kayg commented

2026-11-03 16:21:24 +00:00

Author

Owner

J80d → J136: rebase note. J80d is finished; head abaf410a4 on macos-node-265-s1 (#265 comment 77060).

git rebase --onto origin/macos-node-265-s1 70a5376b2 macos-build-j136 applies all 7 of your commits without conflicts. I trial-ran it locally and did not push it.
Use --onto. A plain rebase onto dev would replay the pre-rebase slice-1 commits and bring back the daemon.rs test-cfg conflict, which I already resolved.
On the combined tree, nix develop -c scripts/macos-build-check gave PASS on the Mac (the release build took 6 min 16 s through the new checking linker wrapper).

**J80d → J136: rebase note.** J80d is finished; head `abaf410a4` on `macos-node-265-s1` (#265 comment 77060). - `git rebase --onto origin/macos-node-265-s1 70a5376b2 macos-build-j136` applies all 7 of your commits **without conflicts**. I trial-ran it locally and did not push it. - Use `--onto`. A plain rebase onto dev would replay the pre-rebase slice-1 commits and bring back the `daemon.rs` test-cfg conflict, which I already resolved. - On the combined tree, `nix develop -c scripts/macos-build-check` gave **PASS** on the Mac (the release build took 6 min 16 s through the new checking linker wrapper).

kayg referenced this issue

2026-11-04 02:26:06 +00:00

macOS mesh node #265

kayg referenced this issue

2026-11-04 02:57:23 +00:00

macOS mesh node #265

kayg referenced this issue

2026-11-04 08:05:02 +00:00

macOS mesh node #265

kayg referenced this issue

2026-11-04 12:52:56 +00:00

macOS mesh node #265