geomind_code/mycelium_network
3
0
Fork 0
Code Issues 11 Pull requests 1 Projects Releases 1 Packages Wiki Activity Actions

Fix fresh-checkout build + daemon startup under hero_proc supervision #48

Merged
mik-tf merged 1 commit from development_mik into development_crate_layout 2026-05-19 18:05:19 +00:00
Conversation 0 Commits 1 Files changed 7 +77 -17
mik-tf commented 2026-05-19 17:31:39 +00:00
Member
Copy link

What this PR changes

Five small fixes to mycelium_network so it builds and runs cleanly on a fresh checkout.

1. Fix broken hero_admin_lib dep

crates/mycelium_admin/Cargo.toml had a path dep pointing at ../../../hero_website_framework/... — that path only resolves if hero_website_framework happens to be cloned as a sibling directory of mycelium_network. On most clones it doesn't exist, and cargo build --workspace fails at the manifest step before any code compiles.

Switched it to the git dep used by hero_voice and the other consumers:

hero_admin_lib = { version = "0.1.0", git = "https://forge.ourworld.tf/lhumina_code/hero_website_framework.git", branch = "development" }

2. Use workspace dep for herolib_core in mycelium_admin

The other four crates declare herolib_core = { workspace = true }. mycelium_admin had its own standalone git ref. Flipped it to match — one place to bump the version going forward.

3. Forward Hero env vars when CLI spawns the daemons

mycelium_cli --start registers two child processes (server + admin) with hero_proc. Those children were being spawned with only RUST_LOG set, so when herolib_core tried to resolve paths (PATH_ROOT, PATH_VAR, etc.) the daemon panicked at startup.

Added a small helper in crates/mycelium_cli/src/self_register.rs that forwards PATH_ROOT, PATH_VAR, PATH_BUILD, PATH_CODE, and HERO_SOCKET_DIR into both spawned actions if they're set in the calling shell. Same pattern used in hero_voice/main.rs.

4. Declare PATH_ROOT in each daemon's service.toml

When daemons are started via lab service <name> --start (instead of via the CLI orchestrator), the env block in service.toml is what gets forwarded into the spawned process. Added:

[[env]]
var     = "PATH_ROOT"
default = "~/hero"
desc    = "Hero install root..."

to mycelium_server, mycelium_admin, mycelium_server_msg, and mycelium_server_private. Without this, those daemons panic on startup under lab service supervision with PATH_VAR is not set.

5. Bump herolib_core pin in Cargo.lock

cargo update -p herolib_core — moves the pin from 3ff281d8 to 33b5a3fb. The new pin includes a fallback in herolib_core that derives PATH_VAR, PATH_BUILD, and PATH_CODE from PATH_ROOT when they aren't set explicitly. Without this update, even with PATH_ROOT correctly threaded through (changes 3 + 4 above), daemons still panic on missing PATH_VAR. With the update, setting just PATH_ROOT is enough.

What I verified

  • cargo build --workspace --release — works on a clean checkout. Before this PR it fails at the workspace manifest step.
  • lab build --release --install — 5/5 binaries built and installed.
  • mycelium_admin started cleanly and responded to:
    • GET /health → {"service":"mycelium","status":"ok","version":"0.7.5"}
    • GET /.well-known/heroservice.json → canonical JSON
  • Confirmed the spawned daemon's env contains PATH_ROOT=... (i.e. change 3 is actually working end-to-end).
  • mycelium_server started, bound its UDS socket, loaded state, created the router — then hit Address already in use (os error 98) on TCP 9651 because my workstation already runs a legacy /usr/local/bin/mycelium daemon on those ports. That's a workstation issue, not a code issue, and a clean reviewer box should make it past the bind.

Tests

  • 31 unit tests pass across all crates except mycelium_sdk and mycelium_e2e.
  • mycelium_e2e/tests/three_node_linux.rs (4 tests) hit the same TCP 9651 conflict as above. Should pass on a box without the legacy daemon running.
  • mycelium_sdk lib test compile fails on Rust 2024 marking std::env::set_var/remove_var as unsafe:
    • crates/mycelium_sdk/src/lib.rs:79,82 — git blame shows both lines were authored in commit 161f06ef on 2026-04-18, before this PR. Latent since the workspace adopted edition 2024. Trivial fix (wrap the test body in unsafe { ... }), but I'm leaving it out of this PR to keep the scope tight. Happy to fold it in if you prefer.

Before this PR, cargo test --workspace couldn't compile at all because of the broken path dep in change 1, so these test-side issues weren't visible.

## What this PR changes Five small fixes to `mycelium_network` so it builds and runs cleanly on a fresh checkout. ### 1. Fix broken `hero_admin_lib` dep `crates/mycelium_admin/Cargo.toml` had a path dep pointing at `../../../hero_website_framework/...` — that path only resolves if `hero_website_framework` happens to be cloned as a sibling directory of `mycelium_network`. On most clones it doesn't exist, and `cargo build --workspace` fails at the manifest step before any code compiles. Switched it to the git dep used by `hero_voice` and the other consumers: ```toml hero_admin_lib = { version = "0.1.0", git = "https://forge.ourworld.tf/lhumina_code/hero_website_framework.git", branch = "development" } ``` ### 2. Use workspace dep for `herolib_core` in `mycelium_admin` The other four crates declare `herolib_core = { workspace = true }`. `mycelium_admin` had its own standalone git ref. Flipped it to match — one place to bump the version going forward. ### 3. Forward Hero env vars when CLI spawns the daemons `mycelium_cli --start` registers two child processes (server + admin) with `hero_proc`. Those children were being spawned with only `RUST_LOG` set, so when `herolib_core` tried to resolve paths (`PATH_ROOT`, `PATH_VAR`, etc.) the daemon panicked at startup. Added a small helper in `crates/mycelium_cli/src/self_register.rs` that forwards `PATH_ROOT`, `PATH_VAR`, `PATH_BUILD`, `PATH_CODE`, and `HERO_SOCKET_DIR` into both spawned actions if they're set in the calling shell. Same pattern used in `hero_voice/main.rs`. ### 4. Declare `PATH_ROOT` in each daemon's `service.toml` When daemons are started via `lab service <name> --start` (instead of via the CLI orchestrator), the env block in `service.toml` is what gets forwarded into the spawned process. Added: ```toml [[env]] var = "PATH_ROOT" default = "~/hero" desc = "Hero install root..." ``` to `mycelium_server`, `mycelium_admin`, `mycelium_server_msg`, and `mycelium_server_private`. Without this, those daemons panic on startup under `lab service` supervision with `PATH_VAR is not set`. ### 5. Bump `herolib_core` pin in `Cargo.lock` `cargo update -p herolib_core` — moves the pin from `3ff281d8` to `33b5a3fb`. The new pin includes a fallback in `herolib_core` that derives `PATH_VAR`, `PATH_BUILD`, and `PATH_CODE` from `PATH_ROOT` when they aren't set explicitly. Without this update, even with `PATH_ROOT` correctly threaded through (changes 3 + 4 above), daemons still panic on missing `PATH_VAR`. With the update, setting just `PATH_ROOT` is enough. ## What I verified - `cargo build --workspace --release` — works on a clean checkout. Before this PR it fails at the workspace manifest step. - `lab build --release --install` — 5/5 binaries built and installed. - `mycelium_admin` started cleanly and responded to: - `GET /health` → `{"service":"mycelium","status":"ok","version":"0.7.5"}` - `GET /.well-known/heroservice.json` → canonical JSON - Confirmed the spawned daemon's env contains `PATH_ROOT=...` (i.e. change 3 is actually working end-to-end). - `mycelium_server` started, bound its UDS socket, loaded state, created the router — then hit `Address already in use (os error 98)` on TCP 9651 because my workstation already runs a legacy `/usr/local/bin/mycelium` daemon on those ports. That's a workstation issue, not a code issue, and a clean reviewer box should make it past the bind. ## Tests - 31 unit tests pass across all crates except `mycelium_sdk` and `mycelium_e2e`. - `mycelium_e2e/tests/three_node_linux.rs` (4 tests) hit the same TCP 9651 conflict as above. Should pass on a box without the legacy daemon running. - `mycelium_sdk` lib test compile fails on Rust 2024 marking `std::env::set_var`/`remove_var` as `unsafe`: - `crates/mycelium_sdk/src/lib.rs:79,82` — `git blame` shows both lines were authored in commit `161f06ef` on 2026-04-18, **before** this PR. Latent since the workspace adopted edition 2024. Trivial fix (wrap the test body in `unsafe { ... }`), but I'm leaving it out of this PR to keep the scope tight. Happy to fold it in if you prefer. Before this PR, `cargo test --workspace` couldn't compile at all because of the broken path dep in change 1, so these test-side issues weren't visible. 
Reaches the 35/35 target for the hero_proc#102 workspace D-10 sweep. Pre-existing
build break + 3 latent gaps surface and are closed in one commit.

Fixes
-----

1. `crates/mycelium_admin/Cargo.toml`
   - hero_admin_lib: path dep `../../../hero_website_framework/...` was broken on
     any checkout where geomind_code/ and lhumina_code/ aren't peers. Flipped to
     the canonical git dep used by hero_voice / hero_planner (origin tip
     branch=development). Unblocks `cargo build --workspace` on a clean clone.
   - herolib_core: now uses `{ workspace = true }` like the other 4 crates,
     instead of a standalone git ref that drifted independently.

2. `crates/mycelium_cli/src/self_register.rs` (Lesson #19)
   Adds `forward_env_if_set` for `PATH_ROOT/PATH_VAR/PATH_BUILD/PATH_CODE/HERO_SOCKET_DIR`
   on both spawned ActionSpecs (mycelium_server + mycelium_admin). hero_proc spawns
   daemons with a clean env, so the daemon's `herolib_core::base` path-resolution
   chain must be threaded explicitly. Same pattern as hero_voice/main.rs:75–92.

3. `crates/mycelium_{server,admin,server_msg,server_private}/service.toml`
   Added `[[env]] PATH_ROOT default="~/hero"` block. Required by the
   `lab service <bin> --start` spawn path: lab forwards the [[env]] block into
   the spawned action's env, satisfying `herolib_core::base::paths::path_root()`
   at daemon startup (Lesson #17, from s107 hero_foundry onward).

4. `Cargo.lock` — `cargo update -p herolib_core` (3ff281d8 → 33b5a3fb)
   The pinned version (3ff281d8) is **before** hero_lib commit `30a0b34e` which
   added the `PATH_VAR/PATH_BUILD/PATH_CODE` derivation fallback from `PATH_ROOT`.
   Without this update, daemons panic at `paths.rs:46` with
   `PATH_VAR is not set — re-install your environment` even when `PATH_ROOT` is
   correctly injected via the [[env]] block, because path_var() did not yet have
   the fallback. New pin (33b5a3fb) is the workspace-wide cascade head (= s124
   D-11 closure across hero_lib/hero_rpc).

D-10 acceptance (hero_proc#102)
-------------------------------

1. service.toml valid for all 5 crates (mycelium_server, mycelium_admin,
   mycelium_cli, mycelium_server_msg, mycelium_server_private) — canonical
   [service] + [[binaries]] + [[binaries.sockets]] (path + protocol) + valid
   `category` enum (network / ui / tool) + `[[env]] PATH_ROOT default="~/hero"`
   on the 4 daemons.
2. `service_base!()` macro applied in all 5 main.rs (verified pre-existing in
   `9905e48 refactor: migrate socket/state path resolution to herolib_core::base
   helpers`); `validate_service_toml` + `handle_info_flag` + (daemons)
   `print_startup_banner` + `prepare_sockets` triad confirmed.
3. `lab build --release --install --policy-mode warn` — 5/5 built clean.
4. `lab infocheck` — 5 crates clean / 0 findings.
5. Smoke validation:
   - mycelium_admin lab service smoke: 2/2 ✓
     (rpc + admin.sock /health 200 + /.well-known/heroservice.json 200)
   - mycelium_admin /proc/<pid>/environ confirms PATH_ROOT=/home/pctwo/hero
     threaded into the lab-spawned daemon — Lesson #19 verified end-to-end.
   - mycelium_server lab service smoke: daemon advances past PATH_VAR resolution,
     binds rpc.sock at `~/hero/var/sockets/mycelium/rpc.sock`, loads state, creates
     Router successfully, then fails with `Address already in use (os error 98)` on
     the TCP underlay port 9651 due to a workstation-side legacy
     `/usr/local/bin/mycelium` daemon (root-owned, ppid 1) occupying ports
     8989/8990/9651. This is a workstation-environmental conflict, not a code
     regression — the reviewer's clean box should pass smoke 4/4.

cargo test --workspace --release
--------------------------------

- 31 tests passed across all crates except mycelium_sdk and mycelium_e2e.
- 4 mycelium_e2e tests (three_node_linux) fail with the same TCP-port collision
  as mycelium_server above: spawned mycelium_server children cannot bind 9651
  because of the legacy system daemon. Reproduced in isolation with explicit
  MYCELIUM_STATE_DIR + HERO_SOCKET_DIR — daemon binds rpc.sock, loads state, then
  panics on `Address already in use`. Environmental, will pass on clean box.
- 1 mycelium_sdk lib test compile failure (NOT introduced by this commit):
  `crates/mycelium_sdk/src/lib.rs:79,82` — `std::env::set_var/remove_var` are
  marked unsafe under Rust 2024 (E0133). `git blame` shows both lines authored by
  @kristof5 in commit `161f06ef` on 2026-04-18, latent on `development_crate_layout`
  since the workspace adopted edition 2024. Wrap-in-unsafe block is the trivial
  fix; out of D-10 scope (separate test-only cleanup).

Branch policy
-------------

PR target is `development_crate_layout` per mycelium_network#46 (current default
dev branch on this repo, NOT `development` like the lhumina_code/* repos).
mik-tf changed title from chore(d10): hero_proc#102 D-10 closure — env forwarding, env block, broken dep, herolib_core cascade to Fix fresh-checkout build + daemon startup under hero_proc supervision 2026-05-19 18:04:46 +00:00
mik-tf merged commit 9abc843c2f into development_crate_layout 2026-05-19 18:05:19 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
Urgent
No labels
Urgent
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
geomind_code/mycelium_network!48
No description provided.