build: complete warm-build hashes.txt coverage (bzImage, tools.cpio, initrd.cpio.xz FORCE; busybox symlinks fix; doc cleanup)

[tlaurion]

Supersedes #2178. Completes the warm-build hashes.txt completeness work started in #2174. Fixes emergency bug introduced by #2174.

Commit 1: build: improve clean target @echo messages and fix reproducibility doc

Makefile clean targets

• Rewrite real.gitclean / gitclean_keep_packages / gitclean_keep_packages_and_build @echo messages to name the git clean -fxd command and describe exact preserve rules
• Tighten real.remove_canary_files-extract_patch_rebuild_what_changed @echo messages: make clear this target deletes .canary stamps and the re-extract/rebuild cascade happens on the next make invocation

doc/reproducible-builds.md verification section

• Split verification into "Same commit: everything should match" vs "Different commits: ROMs always differ" (GIT_HASH in /etc/config causes expected cascade)
• Remove stale real.gitclean_keep_packages prerequisite -- after Fix build reproducibility regression from PR #2136 (musl-cross-make binutils bump) #2174 FORCE fixes, warm builds produce complete hashes.txt
• Remove incorrect warm-build partial-hashes.txt claim
• Add CI hashes.txt download reference with CircleCI artifact URL example
• Add clean working tree prerequisite (dirty tree sets GIT_STATUS=dirty, changing ROM output)
• Restore bzImage in same-commit step-down path
• Upgrade comprehensive diff from hash-only to hash+path comparison

Commit 2: modules/linux: add FORCE and UNCHANGED guard to bzImage rule; fix busybox symlinks

bzImage FORCE + UNCHANGED

• Add FORCE to bzImage copy/hash rule prerequisite -- matching the modules.cpio fix in Fix build reproducibility regression from PR #2136 (musl-cross-make binutils bump) #2174 -- so warm builds always re-run the recipe that records the kernel hash
• Replace direct do-copy-to-target with cmp-based UNCHANGED pattern: copy to .tmp, compare, mv only when changed, print UNCHANGED when identical, always append hash/size to hashes.txt
• When unchanged, preserve mtime to prevent cascading .bundled rebuild

Emergency fix: busybox symlinks broken by #2174

• Fix build reproducibility regression from PR #2136 (musl-cross-make binutils bump) #2174 replaced make install with applets/install.sh --symlinks to avoid re-linking without SOURCE_DATE_EPOCH. However, install.sh reads symlink names from busybox.links, which was not regenerated and is missing after cache restore or make clean. Result: only the bare busybox binary was installed; all 137 applet symlinks ([, ash, cat, cp, ls, ...) were absent from tools.cpio, breaking the initrd.
• Fix: regenerate busybox.links via applets/busybox.mkll before calling install.sh, using HOSTCC=gcc (needed outside busybox's make context) and include/autoconf.h (busybox 1.36.1 uses this, not config.h), with a test -s guard so empty generation fails the build loudly.

Verification (commit c4c59764b64)

Built locally with ./docker_repro.sh make BOARD=EOL_t480-hotp-maximized (warm build, clean GIT_STATUS).

137 busybox symlinks now present in tools.cpio:

$ cpio -itv < tools.cpio | grep -c -- "-> busybox"
137
$ cpio -itv < tools.cpio | grep -- "-> busybox" | head -5
lrwx------   0 root    0 root    7 Dec 31  1969 bin/[ -> busybox
lrwx------   0 root    0 root    7 Dec 31  1969 bin/[[ -> busybox
lrwx------   0 root    0 root    7 Dec 31  1969 bin/ash -> busybox
lrwx------   0 root    0 root    7 Dec 31  1969 bin/cat -> busybox
lrwx------   0 root    0 root    7 Dec 31  1969 bin/cp -> busybox

Local ROM hash: [pending]
CI cross-check: [pending CircleCI pipeline for c4c59764b64]

[Copilot]

Pull request overview

Updates the ROM reproducibility verification documentation to reflect recent Makefile target renames and to clarify what can (and cannot) be expected to match when comparing builds from the same vs different commits.

Changes:

• Removes stale prerequisite guidance (including the renamed real.gitclean_keep_packages target reference and the outdated warm-build hashes.txt claim).
• Splits verification guidance into Same commit (ROM should match) vs Different commits (ROM expected to differ due to embedded GIT_HASH).
• Rephrases the troubleshooting flow to emphasize drilling down from ROM → initrd/bzImage → tools.cpio → individual files.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (1)

doc/reproducible-builds.md:120

• Wording: “Forged example URL” can read as recommending tampering; this looks like a placeholder/example format. Consider renaming to a neutral description so readers don’t misinterpret it.

  Forged example URL:

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Suppressed comments (2)

modules/linux:209

• cmp --quiet "$@.tmp" "$@" will print an error on the first build because $@ doesn’t exist yet. Even though the ! cmp condition will take the "changed" branch, the stderr noise is avoidable by explicitly handling the missing-target case.

	@if ! cmp --quiet "$@.tmp" "$@" ; then \
		mv "$@.tmp" "$@" ; \
		touch "$@" ; \
	else \
		echo "$(DATE) UNCHANGED $(@:$(pwd)/%=%)" ; \

modules/linux:205

• Using do-copy with $@.tmp is problematic here: do-copy runs cmp "$1" "$2" before copying and prints sha256/stat for the destination. Since $@.tmp usually doesn’t exist yet, this will emit cmp: ... No such file noise and log hash/size lines for the temporary file rather than the real target.

This issue also appears on line 205 of the same file.

$(build)/$(BOARD)/$(LINUX_IMAGE_FILE): $(build)/$(linux_dir)/.build FORCE
	$(call do-copy,$(dir $<)/$(linux_output),$@.tmp)
	@if ! cmp --quiet "$@.tmp" "$@" ; then \

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (1)

modules/linux:205

• The bzImage rule copies to $@.tmp via do-copy, but do-copy always prints sha256/stat output for its destination. That makes the build log show hash/size for the temporary file (and it may be deleted on the UNCHANGED path), which is misleading noise in addition to the later tee -a $(HASHES)/$(SIZES) lines for $@. Prefer a plain cp -a (or a helper that doesn’t emit hashes) for the temporary copy.

$(build)/$(BOARD)/$(LINUX_IMAGE_FILE): $(build)/$(linux_dir)/.build FORCE
	$(call do-copy,$(dir $<)/$(linux_output),$@.tmp)
	@if ! cmp --quiet "$@.tmp" "$@" ; then \

[tlaurion]

pushing as bugfix otherwise busybox missing symlinks causes brick