commit a32b9f4c24f82a6b9db70d924dc4bed3777995af
Author: lucascouts <lucascs@protonmail.com>
Date: Thu Aug 27 19:21:45 2026 -0300

add(metadata/{md5-cache/app-admin/awscli2-2.36.33, md5-cache/app-editors/vim-9.2.1015, md5-cache/app-editors/vim-core-9.2.1015, md5-cache/dev-util/codex-desktop-bin-26.820.80927, md5-cache/dev-util/spirv-tools-1.4.357.0_p20260827, md5-cache/sys-apps/asus-ec-sensors-0_p20260827, md5-cache/sys-apps/flatpak-1.18.2}), mod(.autoupdate/packages.toml), up(app-admin/awscli2-2.36.32 -> 2.36.33, app-editors/{vim, vim-core}-9.2.1014 -> 9.2.1015, dev-util/{codex-desktop-bin-26.820.71523 -> 26.820.80927, spirv-tools-1.4.357.0_p20260825 -> 1.4.357.0_p20260827}, sys-apps/{asus-ec-sensors-0_p20260823 -> 0_p20260827, flatpak-1.18.1 -> 1.18.2})

commit ebfb2be457610db6bbaaf8e04f7863af578a3eb5
Author: lucascouts <lucascs@protonmail.com>
Date: Sun Aug 23 13:24:22 2026 -0300

add(metadata/md5-cache/sys-apps/asus-ec-sensors-0_p20260823), mod(.autoupdate/packages.toml), up(sys-apps/asus-ec-sensors-0_p20260809 -> 0_p20260823)

Upstream delta is two lines in asus-ec-sensors.c plus a README row: the DMI
entry for ROG STRIX Z490-A GAMING. The board table goes 60 -> 61, so the
pkg_postinst elog was updated to match -- it names the count, and a stale
count reads as a missing board to anyone diagnosing an -ENODEV modprobe.

No build-system change, so the two pieces of ebuild logic that exist for
this package are untouched: the `modargs=( -C $ M=$ )` that
keeps kbuild from resolving the kernel via `uname -r`, and the
/lib/depmod.d override that wins over the in-tree copy on Linux 7.2+.

Verified: pkgcheck clean (re-run with --cache=no, since a repeat scan can
be silenced by the result cache); `ebuild ... compile` produces a .ko whose
vermagic is 7.1.9-gentoo-dist, matching /usr/src/linux rather than the
running kernel; modinfo lists 61 DMI aliases including
rn*ROGSTRIXZ490-AGAMING*; a full unprivileged `ebuild ... install` yields
an intact image with the depmod.d override. Not merged -- no root on this
host -- so runtime load was not exercised.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

commit e5d6a6f04571e3ae0fd6ddb66f62b40a6d3a102b
Author: lucascouts <lucascs@protonmail.com>
Date: Mon Aug 10 19:59:14 2026 -0300

add(metadata/{md5-cache/app-admin/ansible-core-2.21.3, md5-cache/app-editors/vim-9.2.0933, md5-cache/app-editors/vim-core-9.2.0933, md5-cache/app-editors/zed-1.16.0_pre20260810, md5-cache/app-office/obsidian-1.13.6, md5-cache/dev-util/mesa_clc-26.3.0_pre20260810, md5-cache/dev-util/spirv-tools-1.4.357.0_p20260810, md5-cache/dev-util/vulkan-utility-libraries-1.4.359_p20260810, md5-cache/media-libs/mesa-26.3.0_pre20260810, md5-cache/media-libs/vulkan-layers-1.4.359_p20260810, md5-cache/media-libs/vulkan-loader-1.4.359_p20260810, md5-cache/sci-ml/ik_llama-cpp-0_pre20260810, md5-cache/sci-ml/llama-cpp-0_pre10354, md5-cache/sci-ml/ollama-0.32.7, md5-cache/sci-ml/ollama-bin-0.32.7, md5-cache/sys-apps/asus-ec-sensors-0_p20260809}), del(metadata/{md5-cache/sci-ml/ollama-bin-0.32.5, md5-cache/sys-apps/asus-ec-sensors-0_p20260711-r1}), mod(sys-apps/asus-ec-sensors-0_p20260809), mod(.autoupdate/packages.toml), up(app-admin/ansible-core-2.21.2 -> 2.21.3, app-editors/{vim-9.2.0927 -> 9.2.0933, vim-core-9.2.0927 -> 9.2.0933, zed-1.16.0_pre20260809 -> 1.16.0_pre20260810}, app-office/obsidian-1.13.4 -> 1.13.6, dev-util/{mesa_clc-26.3.0_pre20260809 -> 26.3.0_pre20260810, spirv-tools-1.4.357.0_p20260807 -> 1.4.357.0_p20260810, vulkan-utility-libraries-1.4.359 -> 1.4.359_p20260810}, media-libs/{mesa-26.3.0_pre20260809 -> 26.3.0_pre20260810, vulkan-layers-1.4.359_p20260807 -> 1.4.359_p20260810, vulkan-loader-1.4.359_p20260809 -> 1.4.359_p20260810}, sci-ml/ollama{,-bin}-0.32.6-r1 -> 0.32.7)

commit 913ede1abeeee57a9e7b5ba82022e92113840351
Author: lucascouts <lucascs@protonmail.com>
Date: Sun Aug 9 19:57:23 2026 -0300

del(sys-apps/asus-ec-sensors-0_p20260711-r1)

commit 62d9cf6cd0b1650b6900d64b174cd2d5352428a7
Author: lucascouts <lucascs@protonmail.com>
Date: Sun Aug 9 18:26:55 2026 -0300

add(sys-apps/asus-ec-sensors-0_p20260809), add(metadata/{md5-cache/dev-util/claude-agent-acp-tui-0.10.1, md5-cache/media-libs/vulkan-loader-1.4.359_p20260809}), mod(.autoupdate/packages.toml), up(dev-util/claude-agent-acp-tui-0.10.0 -> 0.10.1, media-libs/vulkan-loader-1.4.359_p20260807 -> 1.4.359_p20260809)

commit c6809efaf04305236bf7ab7cdfb3528cfa2c2150
Author: lucascouts <lucascs@protonmail.com>
Date: Fri Aug 7 07:15:45 2026 -0300

fix(sys-apps/asus-ec-sensors-0_p20260711 -> -r1): build against the selected kernel, not the running one

linux-mod-r1_src_compile pushd's into $ and runs emake there; it does not
invoke `make -C <kernel> M=<src>` itself. GNU make only ever looks for
GNUmakefile, makefile and Makefile -- Kbuild is never a candidate, that
preference belongs to kbuild and only applies once the kernel is the one
entering via M=. So the repo's Makefile won, and it resolves the kernel on
its own:

KVER ?= $(shell uname -r)
KDIR ?= /lib/modules/$(KVER)

The module was therefore compiled against whatever is booted while the eclass
installed it under $, taken from /usr/src/linux, and
_modules_sanity_modversion aborted the merge:

built for '7.1.6-gentoo-dist' while it was meant for '7.1.7-gentoo-dist'

The src_compile comment claimed the opposite of what happens and is replaced.

This stayed latent for as long as `uname -r` matched /usr/src/linux, which is
precisely the condition that stops holding when a new kernel lands: dist-kernel:=
triggers the rebuild before the reboot. Rebooting hides the symptom rather than
fixing it, and while the ABIs happen to agree the failure mode is worse than a
died merge -- a module silently built against the wrong tree.

Driving the kernel's own kbuild fixes it for every kernel rather than for this
pair: the Makefile is bypassed, Kbuild is used as intended, and KERNELRELEASE
comes from the tree actually being built against. KV_OUT_DIR is the build
directory rather than KV_DIR, so out-of-source kernel builds and the eclass's
own extmod-build repointing keep working. net-wireless/mt7927-dkms already
passed modargs this way; this package was the overlay's only exception.

Verified on a host running 7.1.6 with /usr/src/linux at 7.1.7, the exact
divergence that exposed the bug: `USE="-modules-sign" ebuild ... clean install`
completes and the staged .ko reports vermagic 7.1.7-gentoo-dist, where it
reported 7.1.6-gentoo-dist before. Also built clean against 6.18.43 out of tree
(vermagic 6.18.43-gentoo-dist) to confirm neither kernel series is favoured:
the driver's only version guard is unaligned.h below 6.12, MILLI comes from
<linux/units.h>, MODULES_KERNEL_MIN stays at 5.15 and no MODULES_KERNEL_MAX is
introduced. 7.2+, which ships the driver in tree, is still handled by the
depmod override. Runtime modprobe is not covered by either check.

Revbump because the installed module changes.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

commit 25b64699c9ba911a26377a8608fa88f4bf2781ec
Author: lucascouts <lucascs@protonmail.com>
Date: Sun Jul 19 14:45:57 2026 -0300

add(sys-apps/asus-ec-sensors-0_p20260711), add(metadata/md5-cache/sys-apps/asus-ec-sensors-0_p20260711), mod(.autoupdate/packages.toml)