Repository navigation
Shorten the build's runner messages and comments; report every missing firmware pattern - #210
Merged
Merged
Conversation
The stamp machinery in common.sh, the stages, every recipe, the service scripts and s6 run files, the guest checks, the config lists and the patch-set READMEs lose their stacked comments, history and rhetoric. Each recipe drops its two-line "a stage 04 recipe, sourced by..." header. Wrong comments are corrected: bc's shim name, shadow's man pages, the boot-success re-arm file, testctl's key list, and the glibc README's claim that the whole patch directory is fingerprinted. Comments only. bash's declare -f output of every changed script is unchanged and no heredoc body moves, so no recipe or stage function re-fingerprints. Files hashed whole still move their step: the service scripts, run files and sysctl file the late services step, the guest checks the tests step, the allowlists stage 06's rootfs step, and the kernel checker's accepted list its check step alone.
The stamp runner's refusals lose their history ("predates this harness")
and their essay, keeping every phrase step-errexit.sh matches; the
unchrooted escape hatch and stage 05's nested-tree check say what they
mean in a line. The console wrapper the console recipe writes gets one-
line comments, and firmware.list's header a tight format table. Stage
01's collapsed grep continuation is split again.
Fingerprints: the runner functions are excluded from every recipe's, and
stage 01's change leaves declare -f identical. The console recipe's
heredoc and firmware.list are step inputs, so the console step, the
linux-firmware step and every stage 04 step after them rebuild, with
stage 06; stage 05 builds on elfutils, earlier, so the kernel does not.
grep -c exits 1 when it counts nothing, and under the step's pipefail that aborted the recipe at the first such pattern, before its MISSING line and the summary that names them all.
DevomB
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The code changes in
build/that the comment-only cleanup (#206) could not make. Stacked on #206: this branch holds #206's commit plus one of its own, so merge #206 first and this diff shrinks to the second commit.build/lib/common.sh, runner functions only (stage_depends_on,_stamp_stale) andrequire_inside_chroot: the refusals lose their history ("it predates this harness", "the step() that recorded FAILED builds as successful", "under the current harness") and their essay. Every phrasetools/tests/step-errexit.shmatches is kept ("Refusing to resume onto changed inputs", "proves nothing", "stamp records a different fingerprint", "rebuilding this step", "has not completed").build/recipes/console.sh: thekryptik-consolewrapper it writes gets one-line comments (inside the heredoc, so this is code for the stamp).build/config/firmware.list: its header becomes a two-line format table and the Intel and graphics notes one line each. Patterns are unchanged;firmware.shstrips everything from#.build/recipes/firmware.sh: a bug fix that rides with the firmware.list edit, since both re-fingerprint the same steps.n="$(... | grep -c .)"exits 1 when a pattern matches nothing, and under the step'sset -Eeuo pipefailthat aborted the recipe at the first such pattern, before itsMISSINGline and the summary naming them all; it is nowgrep -c . || true.build/stages/01-toolchain.sh: a grep whose\continuation had been collapsed into one line padded with spaces is split again.declare -foutput is byte-identical, so the toolchain identity does not move.build/stages/05-kernel.sh: the nested-tree check names its header in one line instead of "See the header of / this file."Build cost. The runner functions are excluded from every recipe's fingerprint, and stage 01's edit leaves
declare -funchanged. The console heredoc re-fingerprints the console step and the late configuration steps after it.firmware.listis an argument of thelinux-firmwarestep, so that step and every stage 04 step after it rebuild (meson, the Wayland stack, wlroots, dwl, the desktop and configuration steps), then stage 06. Stage 05 chains onelfutils, which comes before both, so the kernel does not rebuild; stage 05's verify-install check reruns alone.firmware.listis the expensive part: drop that file if a comment is not worth the desktop-stack rebuild, or merge this together with a pin bump that rebuilds stage 04 anyway (#173 restarts it).Not done, and why:
boot-check.sh's duplicate seatd check and04-base-system.sh'sstep_failure_hinthistory (open PRs change those files);cryptsetup.sh's no-opbenchmark --helpline andman-db.sh's message (each rebuilds a long tail of steps, the latter the kernel too: worth doing only with a version bump of that package).The first Distro run (36989902525) tests the head before the firmware.sh fix; a second one is dispatched on the new head once it finishes.