algaos/algaos-ebuild-tree/eclass/linux-info.eclass, branch next

linux-info.eclass: Fix building modules when /lib/modules is locked down

2025-06-09T20:27:51Z

/lib/modules/*/build is often a symlink. get_version is usually run in pkg_setup as root, so fully resolve it then in case the unprivileged user doesn't have permission to do it later. Closes: https://bugs.gentoo.org/957187 Signed-off-by: James Le Cuirot

linux-info.eclass: Rework get_version and fix output directory detection

2025-06-05T10:55:08Z

get_version used to detect the output directory under /lib/modules if KBUILD_OUTPUT wasn't explicitly set, but this broke when bug #662772 was fixed. The code is still there but effectively does nothing as OUTPUT_DIR is already used to set KV_OUT_DIR above. Bug: https://bugs.gentoo.org/662772 Signed-off-by: James Le Cuirot

linux-info.eclass: Don't try to read KBUILD_OUTPUT from the Makefile

2025-06-05T10:55:07Z

This code originates from 21 years ago, but I don't think it ever made sense. KBUILD_OUTPUT was never written anywhere until last year. Now it is written to the Makefile in the output directory, which is what we're trying to locate in the first place! Signed-off-by: James Le Cuirot

linux-info.eclass: Pass an empty KBUILD_OUTPUT to make in getfilevar

2025-06-05T10:55:06Z

If KBUILD_OUTPUT is set but doesn't exist, the Makefile will try to create it, which will probably fail due to permissions issues. We don't actually want to write anything here, so there is no need to set it. Signed-off-by: James Le Cuirot

linux-info.eclass: Set the correct ARCH for make in getfilevar

2025-06-05T10:55:05Z

Currently, it unsets ARCH, so it uses the build architecture. This is fine if you have the full kernel sources, which is usually the case. However, an install-extmod-build script was added to Linux fairly recently that installs just the files needed to build out-of-tree modules in a distro-neutral manner. It saves space by only including the arch files for the target architecture. I checked the history, and tc-arch-kernel didn't exist when the unset line was added. If it had, this approach probably would have been taken at the time. Signed-off-by: James Le Cuirot

linux-info.eclass: Drop unused myARCH variable in getfilevar_noexec

2025-06-05T10:55:04Z

This was copied from getfilevar but never used. Signed-off-by: James Le Cuirot

linux-info.eclass: Drop reference to BUILD_FIXES variable

2024-07-08T09:42:07Z

Undocumented, and not used by any ebuilds inheriting this eclass. Signed-off-by: Ulrich Müller

linux-info.eclass: Drop support for EAPI 6

2024-07-08T09:42:06Z

Signed-off-by: Ulrich Müller

linux-info.eclass: Drop temporary fix for powerpc

2024-07-08T09:42:06Z

Apparently there is no reference to a .tmp_gas_check file or to a TOUT variable in Linux sources any more. Bug: https://bugs.gentoo.org/75034#c14 Signed-off-by: Ulrich Müller

linux-info.eclass: respect eselect kernel setting in binpkgs

2024-05-17T12:06:46Z

Commit f51cd5b64c14ddfb83488a12d538c66a4a309376 resets kernel environment variables when binpkgs are merged, this makes sense since we care about the system that the binpkg will be installed on, not the system the binpkg was built on. This introduced Bug 931213 as a side-effect, which will be fixed in a separate commit to linux-mod-r1.eclass. In preparation of that, we document and rename the LINUX_INFO_BINARY_RESET to SKIP_KERNEL_BINPKG_ENV_RESET. However, commit f51cd5b64c14ddfb83488a12d538c66a4a309376 also makes binpkgs always use the running kernel version. This behaviour is surprising and confusing. In principle the 'eselect kernel' setting should still be respected when binpkgs are used. Bug: https://bugs.gentoo.org/926063 Bug: https://bugs.gentoo.org/931213 Signed-off-by: Andrew Ammerlaan