Repository navigation
feat(image): prune ICU from the initrd and the leftover tools from both trees - #489
Merged
jeefy merged 1 commit intoOct 18, 2026
Merged
Conversation
…th trees projectbluefin#485 took the libicu*.so* glob out of common.list because /usr needs libicuuc and libicudata (libarchive -> libxml2 -> ICU). That also put all six ICU libraries back into the initrd, where nothing links libxml2 or libarchive; ICU only reaches it through xfsprogs (extra-fs.bst), and its one user, xfs_scrub, is already pruned. A dependency fix is not possible there, so this is a prune entry. initrd (unpacked, measured against v26.10.1215): - ICU, all six libraries (initrd.list): 40.2 MB - xfsprogs beyond mkfs.xfs, fsck.xfs and xfs_repair (initrd.list): 4.2 MB - projectbluefin#487's leftovers that also apply to the initrd, moved from usr.list to common.list: openssl CLI, cracklib tools and word list, util-linux extras, libgcrypt tools, compile_et/mk_cmds, asn1*, pcre2test, bashbug, psl-make-dafsa, xfs_protofile, zsh/fish completions: 2.6 MB - usr/lib/debug (common.list): 0.9 MB GnuPG (incl. gpgconf, gpg-connect-agent, gpg-error) and lvm2 stay in the initrd. /usr: - libicui18n, libicuio, libicutu, libicutest; they link only each other (common.list): 4.6 MB - libxml2's Python bindings; the Homebrew integration sysext's Python closure has no libxml2, so its dedup is unchanged (usr.list): 1.8 MB - usr/lib/debug (common.list): 0.9 MB Before -> after (v26.10.1215 -> this build): - initrd unpacked: 257,379,689 -> 209,511,738 - *.usr.raw: 215,330,816 -> 211,357,696 - disk UKI: 93,467,568 -> 79,671,296 (.initrd 72,644,608 -> 58,848,256) - netboot UKI: 93,468,080 -> 79,671,808 just check-image against the v26.10.1215 usr.raw passes: 26 /usr paths removed, all from the entries above. The prune-list test now also guards libicuuc and libicudata in common.list and usr.list. Assisted-by: Claude Fable 5.1 via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
jeefy
added
the
full-build
Contributor
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: Needs approval from an approver in each of these files:Approvers can indicate their approval by writing |
![hivecommons-hive[bot]](https://avatars.githubusercontent.com/in/3568013?s=60&v=4){kind=small}
jeefy
added a commit
that referenced
this pull request
Oct 18, 2026
…mpress vmlinuz with zstd (#491) * feat(kernel): drop firmware-less Wi-Fi, Bluetooth and GPU drivers; compress vmlinuz with zstd The image ships no /usr/lib/firmware, so Wi-Fi, Bluetooth and the amdgpu, radeon, nouveau and xe GPU drivers cannot drive the hardware they match. patches/freedesktop-sdk/0009 appends to FSDK's fdsdk-config.sh after 0007: it removes WLAN, CFG80211, MAC80211, BT, DRM_AMDGPU, DRM_RADEON, DRM_NOUVEAU and DRM_XE, and switches KERNEL_GZIP to KERNEL_ZSTD. FSDK's module/enable lines put those drivers in expected-configs, so the block snapshots the expected options that are set, runs olddefconfig, and drops from the check only the ones that went from set to unset (192, listed in the build log; matching a local Kconfig run of 7.2.6). Every other expected option is still checked. CEC_CORE/MEDIA_CEC_RC go too: only amdgpu's display code selected them. Measured against a local build of main (b3e4b0d, same dev keys): usr.raw 211,357,696 -> 187,297,792 (-24.1 MB) disk UKI 79,671,296 -> 75,356,672 (-4.3 MB) netboot UKI 79,671,808 -> 75,357,184 (-4.3 MB) .linux 20,685,824 -> 16,372,736 (gzip -> zstd; cfg80211, mac80211 and the Bluetooth core are no longer built in) Against v26.10.1215 (before #489): usr.raw 215,330,816, disk UKI 93,467,568. Modules gone: 201 files / 23.9 MB (Wi-Fi 160 / 13.0 MB, amd 6.1 MB, xe 1.9 MB, nouveau 1.3 MB, radeon 0.9 MB, Bluetooth 27 / 0.6 MB, cec, five DRM helpers only these drivers used, qmi_helpers). Added: libarc4.ko (was built in via mac80211). Kept: i915, simpledrm, virtio-gpu, hyperv_drm, vmwgfx, vboxvideo, gma500, RFKILL (x86 platform drivers use it), DEBUG_INFO_BTF; sound, media, IIO and input module counts unchanged. ast, mgag200 and bochs were not built before this change and still are not. Verified: 0001/0006/0007/0008/0009 git apply cleanly on the freedesktop-sdk-26.08.2 tag; just validate; just export-image (kernel rebuilt, expected-config check passed); just check-image vs 1215 and main; just dogfood-check (Secure Boot diskless boot, failed=0). Assisted-by: Claude Fable 5.1 via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs(nvidia): nouveau is no longer in the base kernel Patch 0009 stops building nouveau, so no driver holds an NVIDIA GPU when the driver sysext merges. nvidia-load's eviction (`modprobe -r nouveau` only when nouveau holds a GPU, sysfs unbind otherwise) is now a no-op and stays as a guard for a kernel that does build nouveau; update the skill and the helper and unit comments to say so. No behaviour change. Assisted-by: Claude Fable 5.1 via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(kernel): guard zstd vmlinuz to x86_64; update nouveau comments arm64 selects no HAVE_KERNEL_*, so `enable KERNEL_ZSTD` stays unset there and FSDK's expected-configs check would fail the aarch64 kernel build with "Missing KERNEL_ZSTD". Wrap the gzip->zstd switch in an x86_64 case, like 0007 does for the x86-only watchdogs, and test both arches. Reword the modprobe-nvidia.conf comment and its test to say the base kernel no longer builds nouveau. Hive-Fix-Attempt: 1/6 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> --------- Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
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.
Follows up #487 with the "quick win" half of the size review after v26.10.1215. −13.8 MB per UKI (−14.8%), −4.0 MB usr.raw, proven by a full export and
check-imageagainst v26.10.1215: initrd 117 paths removed / 0 added,/usr26 removed / 2 added (the 2 are #486's, not this PR's). No kernel, element-graph oros-basechange.The regression this fixes
#485 removed
usr/lib/*-linux-gnu/libicu*.so*fromfiles/prune/common.listso libarchive→libxml2 could keeplibicuuc/libicudatain/usr.common.listis applied to the initrd too, so all six ICU libraries (40 MB unpacked, 11.8 MB zstd) came back into every UKI — and nothing in the initrd links libxml2 or libarchive. ICU reaches the initrd via FSDKxfsprogs.bst(for the already-prunedxfs_scrub), so it cannot be fixed by dropping a dependency;initrd.listnow prunes it. A test guardslibicuuc/libicudataincommon.list/usr.listonly, so this can't silently recur in/usreither.Also pruned
/usr: the four ICU libraries libxml2 does not need (i18n,io,tu,test;readelfshows they NEED only each other), libxml2's Python bindings (no Python in the base; the homebrew-integration sysext dedups against the unprunedos-filesystemand its closure has no libxml2, so its contents are unchanged),usr/lib/debug(valgrind/gdb symbols; neither ships).cal hexdump look mesg more rev ul, asn1*, mpicalc/hmac256/dumpsexp, compile_et/mk_cmds, bashbug, zsh/fish completions) — moved tocommon.list; and everyxfs_*tool exceptxfs_repair,fsck.xfs,mkfs.xfs(fsck.xfs -fexecs onlyxfs_repair;systemd-growfsuses the ioctl; nothing infiles/,elements/bluefin-server/initrdorscripts/dogfood-*calls the rest). Kept on purpose: gpg/gpgv/gpg-agent/keyboxd/gpgconf, lvm +69-dm-lvm.rules, hwdb,magic.mgc, bsdtar.Every glob was run with the real
prune.shagainst copies of the 1215 trees before the export; predicted savings equal measured..initrdsection.linuxidentical)usr.rawjust validate, pytest 1472, ruff, docs-checks pass. Labelledfull-buildfor the boot-test. The medium/structural items from the review (firmware-less kernel modules, libarchive without libxml2,CONFIG_KERNEL_ZSTD) are deliberately not in this PR.