Mirrors codemirror_bridge.rs: Reflect::get + unchecked_into for the
IIFE object literal window.XtermTerminal, XtermOptions as wasm-bindgen
extern type (TS class survives erasure), TerminalHandle with Drop →
destroy(). WASM-only (pub mod wasm gated on target_arch = wasm32).
Dioxus.toml: register /xterm/terminal.{css,js} in both prod and dev
resource arrays. CSS is emitted as a separate file by vite lib mode
(cssCodeSplit:false still extracts one CSS file for IIFE).
The /admin/system feature was never verified against the WASM frontend
target (dx serve builds src/main.rs with --features web and no server).
The entire src/api/database/ tree unconditionally pulled in server-only
crates (axum, sqlparser, dashmap, tokio, tokio_postgres, futures, regex)
and server-gated helpers (get_current_admin_user, parse_session_token,
get_user_by_token), so the frontend build failed with 43 errors.
Root cause + fixes (mirror the established settings.rs pattern):
- Gate server-only imports behind #[cfg(feature = server)] in
status/system_status/sql_console/schema/tasks/backup. The #[server] macro
strips call sites but NOT use statements, so unresolved imports remained.
- Ungate in main.rs: SystemSnapshot is a shared serde
type referenced by ServerStatus.host on both targets; only the sampler
task/SNAPSHOT static are internally server-gated.
- Gate (pure Axum handler, zero WASM consumers) and the
download_backup Axum handler in backup.rs.
- Move restore_backup's validation preamble (regex/backup_path/fs) inside
the #[cfg(feature = server)] block; gate std::path/chrono::Utc imports.
- Add [[bin]] required-features=[server] for generate_highlight_css so the
syntect-dependent build tool is skipped in web-only builds.
Genuine WASM bugs surfaced once the server bodies stopped masking them:
- system.rs: Closure import path (dioxus::prelude::wasm_bindgen ->
wasm_bindgen, matching write.rs); setTimeout expects i32 not u32;
editor_handle() -> editor_handle.read() (Signal is not Fn in 0.7).
- codemirror_bridge.rs: set_schema extern took &SqlSchema but SqlSchema is
a serde type with no IntoWasmAbi. Changed both set_schema signatures to
&JsValue; call site now serializes via serde-wasm-bindgen::to_value.
- system.rs: signals .set() inside spawn/use_future/onclick closures need
bindings (incl. the *_f rebinding copies); added cfg_attr
allow(unused_mut) on the 4 tab components, matching write.rs convention.
Verified: cargo check (default/server) clean; cargo check --features web
--target wasm32-unknown-unknown clean (0 errors); dx check clean; 411
cargo tests pass.
warn_if_app_base_url_unset() is unconditional—localhost dev also
emits a WARN (per csrf.rs doc). The main.rs side note claimed
localhost stays silent, contradicting the implementation.
Adds src/tiptap_bridge.rs with extern bindings for window.TiptapEditor
module, EditorInstance, EditorOptions (builder), UploadEventJs.
EditorHandle unifies instance + 4 closures lifecycle (Drop calls destroy).
consume_upload_event replaces the old polling body. make_upload_closure
does the /api/upload fetch in Rust (web_sys) instead of eval'd JS.
DB migration failure at startup panicked via .expect() (main.rs:223),
dumping a raw backtrace for what is usually just "PostgreSQL isn't running
yet". Worse, the 1.6s runtime retry budget was reused at startup, so it
almost always failed against a cold/slow DB (docker-compose without
healthchecks, local Postgres not started, etc.).
Changes:
- pool: extract build_pg_config() (returns Result, no panic); add
validate_database_url() so URL-format / DB_POOL_SIZE errors surface as
friendly exit(1) instead of hitting the LazyLock's unreachable .expect()
- pool: add get_conn_for_startup() — a startup-only retry loop with a
configurable total-duration window (MIGRATE_STARTUP_TIMEOUT_SECS,
default 30s, 500ms interval), separate from the runtime get_conn()
anti-avalanche fast retry (untouched)
- migrate: split run() into run_on_conn(&mut conn) so callers control the
connection-acquisition strategy; main.rs pairs this with the startup
retry. Advisory-lock release comment updated: exit(1) terminates the
process just like panic, so the session-level lock is still freed
- main: startup fatal errors (bad URL, DB unreachable, migration failed)
now each emit tracing::error! + eprintln! ERROR + targeted HINT, then
exit(1) — no panic, no backtrace nudge
- .env.example / AGENTS.md: document MIGRATE_STARTUP_TIMEOUT_SECS
No runtime behavior change: get_conn() and its ~40 call sites are
unchanged. Advisory-lock safety preserved (documented in migrate.rs).
upload_image declared ConnectInfo<SocketAddr> as a required Axum extractor,
but dioxus::server::serve() owns the listener and cannot call
into_make_service_with_connect_info::<SocketAddr>(), so the request extension
is absent on manually-merged routes. The extractor failed and Axum returned
500. dev (dx serve) passed by luck; release builds failed consistently.
Match serve_image's graceful-degradation pattern: take
Option<Extension<ConnectInfo<SocketAddr>>>, fall back to the 'unknown'
rate-limit bucket when missing. Production should deploy behind a reverse
proxy with TRUSTED_PROXY_COUNT so rate limiting keys on the real client IP.
Also correct the misleading comment in main.rs: axum 0.8 does have
ConnectInfoLayer/into_make_service_with_connect_info; the real blocker is
that Dioxus owns the listener.
main() is sync, so the async migrate::run() is driven by a dedicated
multi-thread tokio runtime that is built, blocked-on, and dropped before
dioxus::server::serve() starts its own runtime. This keeps the two
runtimes from overlapping.
- COMPRESSION_ALGORITHMS defaults to 'all' when unset
- Support 'none'/'off' to explicitly disable compression
- Add parse_compression_algorithms helper and unit tests
- Update .env.example comment to reflect new default
P0 blockers:
- Fix migration numbering conflict and duplicate indexes
- Change comments.post_id FK to ON DELETE CASCADE
- Restrict public post detail endpoint to published posts only
- Fix rate-limiting IP extraction and fallback to ConnectInfo
- Harden HTML sanitizer: deny unknown URL schemes, restrict data URIs
- Remove session token from login response body
- Enforce image pixel/dimension limits on upload and serving
P1 high-risk:
- Validate uploads by magic bytes and decode GIF/WebP
- Add pagination/rate-limiting to search, tag posts, and comments
- Make first-admin registration and slug uniqueness check atomic
- HTML-escape comment author fields
- Improve HTML minify cache key and skip admin/error responses
- Add mobile navigation menu
P2 accessibility/quality:
- Associate form labels with inputs
- Key PostDetail article by slug to re-init scripts on navigation
- Improve image viewer keyboard accessibility
- Make theme toggle SSR-friendly and add aria-label
- Invalidate slug 404 cache on create and pending count on new comment
- Deduplicate tags case-insensitively
P3 cleanup:
- Remove unused tower-http dependency, expand make clean
- Configure DB pool timeouts and verified recycling
- Run background cleanup tasks immediately on startup
- Use SHA-256 for stable disk cache keys
- Log DB errors with Display instead of Debug
- Update README migration instructions
All tests pass (321), clippy clean, dx check clean.
- Add lol_html-based HTML whitespace minifier (utils/html_minify.rs)
that collapses whitespace and strips comments while preserving
<pre>, <code>, <textarea>, <script>, <style> content
- Add Axum middleware (middleware/minify_html.rs) that minifies
text/html responses with per-URL moka cache (256 entries, 300s TTL)
- Wire middleware into the server router in main.rs
- Add CSS minifier to generate_highlight_css build step
- Apply minify_html to rendered markdown output and TOC
- Add http-body-util dependency for body collection in middleware
- Cache ammonia::Builder with LazyLock (was rebuilt per request)
- Enable tracing release_max_level_info to strip tracing overhead at compile time
- Remove TraceLayer and tower-http trace feature from production
- Increase DB pool size 10→20 (configurable via DB_POOL_SIZE)
- Increase SSR cache TTL 300s→3600s (configurable via SSR_CACHE_SECS)
Benchmark: 7,444 → 9,840 req/s, P99 latency 27.6ms → 11.1ms
