The container-only harness was measuring the wrong thing. Real usage — several
tabs, several boards, clicking into widgets — reaches ~880 MB server and a
650 MB *browser* heap, and the browser was never measured at all.
Profiling build: `docker build --build-arg HOMARR_PROFILING=true . -t
homarr:performance`. React's production bundle strips the hooks DevTools needs,
hence "Profiling not supported". Note reactProductionProfiling in next.config is
read only by the webpack path; under Turbopack it comes from `next build
--profile`, and turbo strips any env var not declared in the task's `env`, so all
three had to line up before the flag actually took effect — verified by markers
present only in React's profiling bundle, and static chunks going 20 -> 84 MB
with source maps.
inspect-container.mts attaches to an already-running container (SIGUSR1 opens
Node's inspector — no rebuild, no restart), reports cgroup/per-process/V8 memory
and can pull a heap snapshot out over CDP. The inspected context has no require
and no dynamic import, so it reads globals and uses CDP domains instead.
analyze-heapsnapshot / analyze-react-fibers / analyze-heap-retainers summarise a
snapshot too large for DevTools to open.
Harness now opens STRESS_BOARDS across STRESS_TABS_PER_BOARD tabs, clicks into
widgets, and records browser heap, DOM nodes and listener count via CDP.
Performance.getMetrics silently returns nothing unless the domain is enabled,
which is why the first run reported a 0 MiB client heap.
* ci: add --affected filtering and fix Turbo cache invalidation
- Remove globalDependencies **/.env which invalidated all caches on any .env change
- Move 40+ env vars from globalEnv to build-specific env (lint/format/typecheck don't need DB_URL etc.)
- Add per-task inputs with test file exclusions so test changes don't trigger build/typecheck
- Enable affectedUsingTaskInputs for task-level --affected detection
- Add --affected to all turbo commands in CI (lint, format, typecheck, build)
- Add fetch-depth: 0 for proper git history comparison
- Add Docker Buildx + GHA layer caching for E2E builds
* ci: add Cloudflare remote cache for Turborepo
Configure turborepo-remote-cache-cloudflare as the remote cache backend.
This commit can be reverted independently to disable remote caching.
* fix(ci): set TURBO_SCM_BASE for --affected to work in PRs
Turbo defaults to comparing against `main`, but PRs target `dev`.
Without explicit base/head refs, --affected falls back to running
all tasks. Use the PR base SHA for pull_requests, the before SHA
for pushes, and HEAD~1 as a last resort.
* perf(typecheck): exclude bundled .cjs/.css from tsc
websocket (835K lines) and tasks (755K lines) had bundled output
files being typechecked due to `include: ["."]` in their tsconfig.
Full typecheck drops from ~70s to ~26s.
* perf(turbo): scope topo task inputs to package.json only
The topo virtual task had no explicit inputs, defaulting to all files.
This caused any source change to cascade cache invalidation through
all dependent packages via ^topo. Since topo is only for dependency
ordering, its hash should only change when dependencies change.
* fix(turbo): disable remote cache signature verification
The cache lives on a private CF worker, signature verification
just adds noise without security benefit.
* fix(turbo): remove runtime env vars from build cache key
These env vars (DB_*, AUTH_*, DOCKER_*, etc.) are all runtime
configuration provided by users when running the Docker container.
They don't influence the build output since Next.js standalone mode
reads process.env at runtime and SKIP_ENV_VALIDATION=true skips
validation during build. Listing them in build.env caused unnecessary
cache busting in local dev.
Also restores coverage report comments per review feedback.
* fix(turbo): include test files in typecheck cache inputs
Test files must be typechecked — excluding them from typecheck inputs
caused changed tests to skip type validation.
* fix: skip env validation in dockerfile not working
* fix: skip env validation in dockerfile not working
* fix: remove linux/arm/v7 for now as node commands are run forever
* fix: remove cache because it is to big?
* refactor: remove redis log transport during build
* fix: add more checks for conditional redis connection
* fix: docker build not working locally
* refactor: move base image to debian
* chore: add arm v7 platform support
* fix: remove armv7 support as not supported by turbo
* chore: profile amd64 build
* chore: enable webpack logging
* chore: disable linux/arm64 for now
* chore: remove profiling from build in dockerfile
* chore: revert to node alpine image
* wip: sso
* feat: add ldap client and provider
* feat: implement login form
* feat: finish sso
* fix: lint and format issue
* chore: address pull request feedback
* fix: build not working
* fix: oidc is redirected to internal docker container hostname
* fix: build not working
* refactor: migrate to ldapts
* fix: format and frozen lock file
* fix: deepsource issues
* fix: unit tests for ldap authorization not working
* refactor: remove unnecessary args from dockerfile
* chore: address pull request feedback
* fix: use console instead of logger in auth env.mjs
* fix: default value for auth provider of wrong type
* fix: broken lock file
* fix: format issue
* config: add dockerfile
* feat: add working dockerfile for nest and next app with sqlite migration
---------
Co-authored-by: Meier Lukas <meierschlumpf@gmail.com>
Thomas Camlong
Ashton BattyandMeier Lukas
homarr-renovate[bot]andMeier Lukas
ManuelandMeier Lukas