docs(DEVELOPER.md): update to catch up the current behavior

[spacewander]

make build-venv only setup the venv. It doesn't build the kong.

[ADD-SP]

overall LGTM, just have a small question.

[ADD-SP]

Thanks you!

CI is red

[ADD-SP]

Suggested change

	dev: install-rust-toolchain build-venv build-openresty install-dev-rocks bin/grpcurl bin/h2client
	dev: install-rust-toolchain build-venv install-dev-rocks bin/grpcurl bin/h2client

This fixed the CI failures. Did the missing build-openresty (as the deps of the dev target) cause issues while developing locally?

Background

The target build-openresty is only for local nginx module/patch development, so there are some assumptions in the Bazel script.

In simple words, the Makefile target invokes the Bazel target @openresty//:dev-just-make to rebuild the OpenResty, but this Bazel target tries to reuse the current Bazel working directory.

kong/build/openresty/BUILD.openresty.bazel

Line 369 in 5885ea4

pushd $(RULEDIR)/openresty.build_tmpdir >/dev/null

The Bazel working directory is not cached across the CI jobs, only the Bazel output directory is cached, so the Makefile target build-openresty is failing on CI.

[spacewander]

This fixed the CI failures. Did the missing build-openresty (as the deps of the dev target) cause issues while developing locally?

The build-openresty builds the OpenResty, which is required by Kong. Is there another way to provide the nginx binary? I take a look at the CI, and it uses bazel build //build:install-openresty under the hood in the build job: https://github.com/Kong/kong/actions/runs/15016755778/job/42253855998?pr=14478

The Makefile task build-openresty will call //build:dev-make-openresty instead of //build:install-openresty, which causes the CI to fail. Is there a difference between an incr build and a full build? I am not so sure...

[ADD-SP]

Is there another way to provide the nginx binary? I

The chain is complicated. In short, the make build-venv eventually builds the OpenResty binary.

[ADD-SP]

Is make build-venv not generating OpenResty binary on dev box? If so, this should be a bug and need to be fixed in another PR.

[ADD-SP]

Is there a difference between an incr build and a full build?

I didn't fully deep dive it.

There are some hacks inside the //build:dev-make-openresty, it rm the binary in the working directory and re-run the make && make install under the OpenResty source.

And for the full build, at least, it doesn't re-run the configure script.

[spacewander]

Is make build-venv not generating OpenResty binary on dev box? If so, this should be a bug and need to be fixed in another PR.

Thanks for your notification. Now I know why:

kong/Makefile

Line 100 in 5885ea4

@if [ ! -e bazel-bin/build/$(BUILD_NAME)-venv.sh ]; then \

the build-venv will check the generated file before running the actual command. When running for the first time, the command generated the venv.sh file, but run failed later when installing luarocks. (because of missing libyaml). When I rerun the command after the fix, the build-venv doesn't nothing, only venv is generated, so I thought this task only generates the venv.