Repository navigation
Conversation
Port the elastos-logger crate authored in 27d85c6 (origin/feat/dkms-esp-port) onto the collaboration base as its own track: Level (Trace<Info<Warn<Error<Critical, "debug" alias), LogRecord + format_line, the LogSink writer seam (Stderr/Stdout/File/Vec sinks), the process-global Logger with pre-init stderr fallback, resolve_level (CLI > env chain > default), the fp() privacy fingerprint, and the lazy log_* macros. Registered as a workspace member in elastos/Cargo.toml. Crate files are ported verbatim; the source commit's PR #15 wirings (dkms-authority, dkms-carrier-node, providers) do not transfer to this base and are deliberately out of scope for this slice. The plan doc the source commit referenced (docs/superpowers/plans/2026-08-17-elastos-logger.md) does not exist in any tree; the reference is dropped — it only ever lived in the commit message, not in the ported files. Verification (from elastos/): - cargo test -p elastos-logger: 15 passed, 0 failed - cargo clippy -p elastos-logger --all-targets -- -D warnings: clean - cargo fmt --all -- --check: clean - git diff --check: clean
… macro arm) The Logger holds one process-wide component; a multi-surface binary like elastos-server needs per-surface names on a single global. Add Logger::emit_as / global emit_as stamping a caller-supplied component, factor the pre-init stderr fallback, and teach log_at! (and thus all log_*! macros) an optional leading 'component: <expr>' argument. Verification: (cd elastos && cargo test -p elastos-logger) — 16 pass; clippy + fmt clean.
Install the process-global logger at startup: threshold resolves CLI --log-level (new global clap arg) > ELASTOS_LOG > Info; sink is a SuppressableStderrSink honoring the same interactive-TUI suppression flag as the existing tracing ConditionalStderr. The tracing subscriber stays in place until all call sites are converted (follow-up slices). Verification: (cd elastos && cargo test -p elastos-server) — 1419 pass, 2 pre-existing env-dependent failures in room::direct (fail identically on the parent commit); fmt + clippy clean.
Convert all ~305 tracing::error!/warn!/info!/debug!/trace! call sites across elastos-server to log_error!/log_warn!/log_info!/log_trace! (debug maps to Trace — the logger has no Debug level), each stamped with a flat per-surface component via log_*!(component: LOG_COMPONENT, ...): gw, auth, wallet, browser, home, effects, infra, bridge, carrier, collab, runtime, vm, content, supervisor, serve, run, update, ipfs, binaries, operator; main.rs logs under the default 'elastos' component. Structured tracing fields are folded into the format strings with message text preserved; sensitive values at INFO/TRACE (session token, node/peer DIDs, principal id, browser exit-stream target) are fingerprinted with fp(). The tracing_subscriber init in main.rs stays until the remaining tracing-emitting crates are converted (later slices), so their events are not silently dropped in the interim; tracing deps drop at the end of the rollout. Verification: (cd elastos && cargo test -p elastos-server) — 1419 pass, 2 pre-existing env-dependent room::direct failures (identical on parent commits); clippy --all-targets and fmt clean; zero tracing:: call sites remain outside the subscriber init.
The vm bridge logged the entire serialized resource request (including write content) at INFO on connection reuse, and the full init config JSON at INFO on connect. Payload content can carry user file data and manifest-provided config values, which violates the secret-free logging rule. Demote the reuse line to TRACE with op name + fp() of the payload for correlation, and log the init config as fp() only. Verification: (cd elastos && cargo clippy -p elastos-server && cargo fmt --all -- --check) clean; behavior covered by existing vm bridge tests in the suite run for the parent commit.
…ogger Triage of the ~77 eprintln! sites in elastos-server: bracketed daemon-style diagnostics convert to the logger, interactive CLI output stays on stderr untouched. Converted: gateway_cmd (14 sites -> gw), serve_cmd (6 -> serve), supervisor (8 -> supervisor), run_cmd (4 -> run), chat cleanup warnings (2 -> chat), share head-publish warnings (3 -> share), host_lock supersession fatal (1 -> lock, log_critical). The '[prefix]' tags drop from messages since the component now carries that context. Kept as eprintln/println (user-facing): interactive chat UI output, runtime_control notices (own runtime_notices_enabled() gate), channel revocation warnings and shutdown notice in main.rs, CLI progress formatting in setup/update/publish/node_cmd, share signing-key display, cfg(test) smoke-skip notices, and all println! command output. Verification: (cd elastos && cargo test -p elastos-server) — 1419 pass, 2 pre-existing env-dependent room::direct failures; clippy --all-targets and fmt clean; git diff --check clean.
Convert all remaining tracing call sites in the workspace's library crates to log_*! macros with flat component names, and drop each crate's tracing dependency: elastos-runtime (54 sites -> runtime), elastos-crosvm (28 -> crosvm), elastos-vz (8 + supervisor bin eprintln -> vz), elastos-storage (22 -> storage), elastos-tls (14 -> tls), elastos-namespace (7 -> ns), elastos-identity (1 -> identity), elastos-guest (1 diagnostic -> guest, cfg-gated so WASI capsule builds stay minimal), elastos-compute (unused tracing dep dropped). Session tokens formerly logged as 8-char prefixes are now fp() fingerprints (elastos-runtime session registry, crosvm provider). The vz concurrent_launch test's tracing_subscriber capture Layer is replaced by an elastos-logger VecSink capture keyed on the vm_console message prefix. elastos-guest's exported log!/log_error! guest-SDK macros and the vz supervisor's VZ_MEDIA_DIAGNOSTIC_SCHEMA JSON emitter are deliberately untouched (public API / machine-readable contract). Verification: (cd elastos && cargo test -p <each converted crate>) all pass incl. elastos-vz concurrent_launch; cargo check/clippy --workspace --all-targets zero warnings; fmt clean; grep shows zero tracing references outside elastos-server's subscriber (removed next).
All tracing call sites across the workspace are converted, so remove the tracing_subscriber init from main.rs along with the ConditionalStderr MakeWriter it fed, and drop the tracing + tracing-subscriber dependencies from elastos-server. The SuppressableStderrSink and set_logging_suppressed() stay — the interactive-TUI suppression flag now gates only the elastos-logger sink. Remaining 'tracing' entries in Cargo.lock are transitive (axum/tower internals), not direct dependencies. Verification: (cd elastos && cargo test -p elastos-server) — 1419 pass, 2 pre-existing env-dependent room::direct failures; cargo clippy --workspace --all-targets zero warnings; fmt clean.
Rename the elastos binary's surface components into hierarchy buckets:
gateway.{http,auth,browser,home,effects,content,bridge,infra,capsule},
cmd.{chat,ipfs,run,serve,share,update}, vm.{crosvm,vz,provider,
supervisor}, host.{binaries,lock,operator}. carrier, collab, and the
library crates (runtime, storage, tls, ns, identity, guest) stay
top-level components.
Verification: (cd elastos && cargo check --workspace --all-targets)
clean; fmt clean; elastos-vz + elastos-logger tests pass (component
strings do not participate in any test assertion except the vm_console
message prefix, which is unchanged).
One more impl LogSink, no Logger change (as the crate's sink NOTE anticipated): JsonRingSink appends one JSON object per record — ts, level, component, target, msg — to <component>.jsonl under a configured directory, rotating to .jsonl.1 at 5MB so each component holds at most ~10MB across two generations. Component names are sanitized to [A-Za-z0-9._-] with leading dots stripped, so hostile names cannot escape the directory. Messages are the exact rendered text the stderr sink sees (already fp()'d) — the sink adds no data of its own. The elastos binary enables it observe-only via ELASTOS_LOG_JSON_DIR; an unusable directory logs a WARN and falls back to stderr-only. Adds serde_json to elastos-logger and a crate README documenting levels, components, the fp() rule, and the JSON format contract for consumers. Verification: (cd elastos && cargo test -p elastos-logger) — 20 pass (4 new: shape, per-component split, rotation bound, hostile-name sanitization); cargo test -p elastos-server — 1419 pass, 2 pre-existing env-dependent room::direct failures; clippy --workspace --all-targets zero warnings; fmt clean.
irzhywau
force-pushed
the
feat/elastos-logger
branch
from
August 25, 2026 07:53
8d05169 to
2e843e1
Compare
…! macros
Add elastos_logger::component!("<name>") and component_mod!(vis name,
"<name>"): they generate per-component trace!/info!/warn!/error!/
critical! macros so call sites carry no component at all. Each crate
declares its logger once in src/logger.rs — library crates via
component! (files then 'use crate::logger;'), the elastos binary via
one component_mod! per surface (files 'use
crate::logger::gateway_auth as logger;') — and logs with
logger::warn!("...{}", x). The generated macros delegate to
log_*!(component: ...), keeping lazy formatting and the call site's
module_path!() target; the log_*! forms remain for the process-default
component (main.rs) and dynamic components.
Macro-hygiene notes baked into the design: the generated macros use
unique internal names re-exported 'as warn' etc. (a macro literally
named warn cannot be use'd — ambiguous with the built-in #[warn]
attribute), the re-export is pub(crate) (textual macros cannot be
re-exported across crates, E0364), and each target declares its own
logger module (main.rs 'mod logger;' reuses the same source file; the
vz supervisor bin declares an inline one).
Migrated all 59 per-file const LOG_COMPONENT sites plus main.rs; the
consts and per-file elastos_logger macro imports are gone (only fp
imports remain where messages fingerprint values). README documents the
new shape.
Verification: (cd elastos && cargo test -p elastos-logger) — 20 pass +
doctest (generated macros through the global, component! and
component_mod! forms); all converted library crate tests pass; cargo
test -p elastos-server — 1419 pass, 2 pre-existing env-dependent
room::direct failures; clippy --workspace --all-targets zero warnings;
fmt clean.
Collaborator
Author
|
Status update (2026-08-25) — where this track stands. Landed (pushed)Crate (slice 1):
Gateway adoption (slice 2):
Library crates (slice 3,
Component naming + AI sink:
Local, pending push
Verification state
Remaining scope (separate slices)
|
Add the logging contract to the places developers and agents actually look: AGENTS.md gains a Logging Discipline section (logger modules at call sites, component buckets, println reserved for CLI UX, secret-free TRACE via fp()); DEBUG.md documents raising verbosity (--log-level, ELASTOS_LOG), the line format, component grepping, and the ELASTOS_LOG_JSON_DIR insight stream; docs/ARCHITECTURE.md gains a Logging and observability section; docs/COMMAND_MATRIX.md records --log-level as a global flag (rule 9); elastos/README.md and the docs index link the crate README. scripts/home-smoke.sh switches its quiet-run env from the now-inert RUST_LOG to ELASTOS_LOG. The smoke's first check fails identically before and after this change on this branch (the Home header renders 'Home CLI / Home', not the expected 'ElastOS Home') — pre-existing drift, left untouched like the room::direct test failures. Verification: git diff --check clean; docs-only plus one env var.
elastos-guest now (target-gated) depends on elastos-logger, so home-cli's lockfile gains the logger's dependency entries (sha2, hex, block-buffer, ...). Regenerated by cargo during the home-smoke run; other capsule lockfiles will refresh the same way on their next build. Verification: cargo build side-effect only; lockfile diff is additive dependency entries.
…endencies Added support for additional targets in rust-toolchain.toml. Updated Cargo.lock files across multiple capsules to include new dependencies such as 'deranged', 'elastos-logger', 'num-conv', 'powerfmt', and 'time' along with their respective transitive dependencies. This ensures compatibility with the latest changes in the project.
Collaborator
Author
|
This PR is potentially set to be abandoned in favor to |
This was referenced Sep 29, 2026
Contributor
|
@irzhywau retire this in favour of tracing, or rescope it? |
6 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
elastos-loggercrate from the tip of PR dKMS/dDRM foundation on the ESP substrate (inspired from flint-5.0) #15 (commit 27d85c6) as its own PR track, per review feedback that the logger is separate work from the protected-content extractionelastos/crates/elastos-logger:Level(case-insensitive parse, aliases, env fallback),LogRecord+format_line(),LogSinktrait (Send+Sync, non-panicking, drop-on-fail) withStderrSink/StdoutSink/FileSink/VecSink,LoggerwithOnceLockglobal and safe pre-init stderr fallback,resolve_levelprecedence (CLI > env chain > default),fp()SHA-256 privacy fingerprint, lazylog_trace!..log_critical!macros, and per-record component override (component:/emit_as)elastos-serverbootstraps the global logger withELASTOS_LOGas the env root, all tracing call sites converted, daemoneprintln!diagnostics routed through the logger, tracing subscriber and dependencies removed; privacy fix so full VM provider payloads are no longer loggedelastos/workspace):elastos-crosvm,elastos-guest,elastos-identity,elastos-namespace,elastos-runtime,elastos-storage,elastos-tls,elastos-vzconverted from tracinggateway.*,cmd.*,vm.*,host.*)JsonRingSink: bounded per-component JSON ring files (enabled viaELASTOS_LOG_JSON_DIR), observe-only and secret-free by constructionOut of scope here (follow-up slices): provider capsule adoption, JS/browser mirror (pending product decision in the PR #15 thread), and any wiring inside the protected-content extraction branches.
Test plan
cd elastos && cargo test -p elastos-logger— 20 tests + doctest passcargo test -p elastos-server— 1419 pass, 2 pre-existing env-dependentroom::directfailurescargo clippy --workspace --all-targetszero warnings;cargo fmt --all -- --checkclean