script/loadtest [duration [clients]] (make loadtest, defaults 10s and 4)
measures pixad in three scenarios: hit (one cached image), miss (a new
source image every request) and herd (each new source image asked for
by all clients at once). For each it prints vegeta's report (requests
per second, p50/p95/p99 latency, status codes), pixad's peak resident
memory and how many requests reached the origin. It is a benchmark, not
run by script/check. The origin it uses, internal/loadtestorigin behind
cmd/loadtest-origin, answers every path with one generated JPEG. Bad
arguments are refused before the build. README.md says how to run and
read it and keeps 1-5k r/s as a target not yet measured at scale.
Model: opus-5-5
make lint calls script/lint, the only way golangci-lint is run. Inside a
container it runs the linter; anywhere else it builds Dockerfile.lint,
whose last step runs script/lint again. Both Dockerfiles set
container=docker to mark the container, since /.dockerenv is missing in
build steps and present on hosts that are themselves containers. The
Dockerfile lint stage runs make lint.
A new CACHEBUST build-arg on every run keeps the lint step from being
served from cache; a tmpfs mount keeps Go's and golangci-lint's caches
out of that step's layer, so runs do not pile up build cache.
script/bootstrap and the nix-shell package lists no longer carry
golangci-lint. golangci-lint config verify is not run: it fetches its
schema over an unpinned live HTTPS call.
Model: opus-4-8 (implementation); opus-5-5 (rework)
The runtime stage declares a HEALTHCHECK that probes
/.well-known/healthcheck.json with busybox wget. script/docker-smoke
(make docker-smoke) builds the image with script/docker, starts it with
a random PIXA_SIGNING_KEY, and passes only once Docker reports the
container healthy within 30 seconds; the container is removed on exit
and its log printed on failure. The Gitea workflow runs it after
script/cibuild; it is not part of make check.
It waits on Docker's health status instead of polling a published host
port because the Gitea job runs in its own container on its own
network, where such a port is not reachable at localhost.
Model: opus-5-5
- Add config validation: signing_key required, minimum 32 characters
- Server now fails to start without valid signing_key (no more runtime errors)
- Add config.example.yml with default whitelist hosts
- Copy config to /etc/pixa/config.yml in Docker image
- Update entrypoint to use --config /etc/pixa/config.yml
- Add config.dev.yml for local Docker development
- Mount dev config in make devserver
- Replace gen2brain/avif, gen2brain/webp, disintegration/imaging with govips
- govips uses libvips via CGO for fast native image processing
- Add libheif-dev to Dockerfile for AVIF support
- Add docker-test Makefile target for running tests in Docker
- Update processor.go to use vips API for decode, resize, encode
- Add TestMain to initialize/shutdown vips in tests
- Remove WASM-based libraries (gen2brain) in favor of native codecs
Performance improvement: AVIF encoding now uses native libheif instead of
WASM, significantly reducing encoding time for large images.
- Build stage: golang:1.24-alpine with vips-dev for CGO image libs
- Runtime stage: alpine:3.21 with vips runtime only
- Pass VERSION build arg for ldflags embedding
- Add 'make docker' target to build image with git version
Hot cache entries now store all data needed to serve a cache hit
without any database access:
- OutputHash (for file lookup)
- ContentType (for Content-Type header)
- SizeBytes (for Content-Length header)
Previously hot cache only stored OutputHash, causing empty
Content-Type headers on cached WebP responses.
- check: default target, runs fmt-check, lint, and test
- fmt-check: verifies code is properly formatted
- fmt: formats code with gofmt
- lint: runs golangci-lint
- test: runs go test
- build: builds pixad binary with version info
- clean: removes build artifacts
clawbot
sneak