Support reproducible builds (except packages)

See docker-library/official-images issue 16044

- `ARG SOURCE_DATE_EPOCH` is added.
  The argument value is consumed by the build scripts to make the binary reproducible.

- `/tmp/*` is removed as they contain files created by `memcached`

- For Debian, `/var/log/*` is removed as they contain timestamps

- For Debian, `/var/cache/ldconfig/aux-cache` is removed as they contain inode numbers, etc.

- For Alpine, virtual package versions are pinned to "0" to eliminate
  the timestamp-based version numbers that appear in `/etc/apk/world` and `/lib/apk/db/installed`

> [!NOTE]
> The following topics are NOT covered by this commit:
>
> - To reproduce file timestamps in layers, BuildKit has to be executed with
>   `--output type=<TYPE>,rewrite-timestamp=true`.
>   Needs BuildKit v0.13 or later.
>
> - To reproduce the base image by the hash, reproducers may:
>   - modify the `FROM` instruction in Dockerfile manually
>   - or, use the `CONVERT` action of source policies to replace the base image.
>     <https://github.com/moby/buildkit/blob/v0.13.2/docs/build-repro.md>
>
> - To reproduce packages, see the `RUN` instruction hook proposed in
>   moby/buildkit issue 4576

Signed-off-by: Akihiro Suda <[email protected]>

Author: AkihiroSuda

1/alpine/Dockerfile

@@ -4,7 +4,11 @@ FROM alpine:{{ .alpine.version }}
 FROM debian:{{ .debian.version }}-slim
 {{ ) end -}}
 
+# The global SOURCE_DATE_EPOCH is consumed by commands that are not associated with a source artifact
+ARG SOURCE_DATE_EPOCH
+
 # add our user and group first to make sure their IDs get assigned consistently, regardless of whatever dependencies get added
+# On Debian, useradd recognizes SOURCE_DATE_EPOCH to reproduce the "lastchanged" field in /etc/shadow.
 {{ if env.variant == "alpine" then ( -}}
 RUN set -eux; \
 	addgroup -g 11211 memcache; \
@@ -24,7 +28,9 @@ RUN set -eux; \
 	apt-get install -y --no-install-recommends \
 		libsasl2-modules \
 	; \
-	rm -rf /var/lib/apt/lists/*
+	rm -rf /var/lib/apt/lists/*; \
+# clean up for reproducibility
+	rm -rf /var/log/* /var/cache/ldconfig/aux-cache
 {{ ) end -}}
 
 ENV MEMCACHED_VERSION {{ .version }}
@@ -34,7 +40,7 @@ ENV MEMCACHED_SHA1 {{ .sha1 }}
 RUN set -eux; \
 	\
 {{ if env.variant == "alpine" then ( -}}
-	apk add --no-cache --virtual .build-deps \
+	apk add --no-cache --virtual .build-deps=0 \
 		ca-certificates \
 		coreutils \
 		cyrus-sasl-dev \
@@ -67,6 +73,8 @@ RUN set -eux; \
 		wget \
 	; \
 	rm -rf /var/lib/apt/lists/*; \
+# clean up for reproducibility
+	rm -rf /var/log/* /var/cache/ldconfig/aux-cache; \
 {{ ) end -}}
 	\
 	wget -O memcached.tar.gz "$MEMCACHED_URL"; \
@@ -78,6 +86,10 @@ RUN set -eux; \
 	cd /usr/src/memcached; \
 	\
 	gnuArch="$(dpkg-architecture --query DEB_BUILD_GNU_TYPE)"; \
+	SOURCE_DATE_EPOCH="$(find . -type f -exec stat -c '%Y' {} + | sort -nr | head -n1)"; \
+	export SOURCE_DATE_EPOCH; \
+# for logging validation/edification
+	date --date "@$SOURCE_DATE_EPOCH" --rfc-2822; \
 	./configure \
 		--build="$gnuArch" \
 		--enable-extstore \
@@ -109,7 +121,7 @@ RUN set -eux; \
 			| sort -u \
 			| awk 'system("[ -e /usr/local/lib/" $1 " ]") == 0 { next } { print "so:" $1 }' \
 	)"; \
-	apk add --no-network --virtual .memcached-rundeps $runDeps; \
+	apk add --no-network --virtual .memcached-rundeps=0 $runDeps; \
 	apk del --no-network .build-deps; \
 {{ ) else ( -}}
 	apt-mark auto '.*' > /dev/null; \
@@ -123,9 +135,13 @@ RUN set -eux; \
 		| xargs -r apt-mark manual \
 	; \
 	apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false; \
+# clean up for reproducibility
+	rm -rf /var/log/* /var/cache/ldconfig/aux-cache; \
 {{ ) end -}}
 	\
-	memcached -V
+	memcached -V ;\
+# clean up for reproducibility
+	rm -rf /tmp/*
 
 COPY docker-entrypoint.sh /usr/local/bin/
 RUN ln -s usr/local/bin/docker-entrypoint.sh /entrypoint.sh # backwards compat