Created attachment 897239 [details] foo-1.2.3.ebuild I'm attempting to port Chromium to use the llvm-r1 eclass after running up against some of the limitations of the older version. I'm trying to build with only testing LLVM/Clang to rule out some weird issues I'm seeing on my dev laptop, however I am unable to do so without setting the USE_EXPAND for the testing slot. This seems unintuitive based on the eclass docs: > The eclass sets IUSE and REQUIRED_USE. The flag corresponding > to the newest supported stable LLVM slot (or the newest testing, > if no stable slots are supported) is enabled by default. Am I holding the eclass wrong? Minimum reproducer ebuild attached. ``` $ doas ebuild chromium-127.0.6533.26.ebuild clean merge ; eix clang; eix llvm; emerge -pv llvm:18 The following REQUIRED_USE flag constraints are unsatisfied: ^^ ( llvm_slot_18 llvm_slot_19 ) The above constraints are a subset of the following complete expression: !headless? ( any-of ( X wayland ) ) official? ( lto ) pgo? ( X !wayland ) qt6? ( qt5 ) screencast? ( wayland ) !system-toolchain? ( libcxx ) ffmpeg-chromium? ( bindist proprietary-codecs ) exactly-one-of ( llvm_slot_18 llvm_slot_19 ) [I] sys-devel/clang Available versions: (15) 15.0.7-r3(15/15g1)^t (16) 16.0.6(16/16)^t (17) 17.0.6(17/17)^t (18) (~)18.1.8(18/18.1)^t (19) **19.0.0.9999(19/19.0.0.9999)*l^t {(+)debug doc +extra ieee-long-double +pie +static-analyzer test verify-sig xml ABI_MIPS="n32 n64 o32" ABI_S390="32 64" ABI_X86="32 64 x32" LLVM_TARGETS="AArch64 AMDGPU ARC ARM AVR BPF CSKY DirectX Hexagon Lanai LoongArch M68k MSP430 Mips NVPTX PowerPC RISCV SPIRV Sparc SystemZ VE WebAssembly X86 XCore Xtensa" PYTHON_SINGLE_TARGET="python3_10 python3_11 python3_12 python3_13"} Installed versions: 17.0.6(17/17)^t(06:47:40 22/06/24)(extra pie static-analyzer xml -debug -doc -ieee-long-double -test -verify-sig ABI_MIPS="-n32 -n64 -o32" ABI_S390="-32 -64" ABI_X86="64 -32 -x32" LLVM_TARGETS="AArch64 AMDGPU ARM AVR BPF Hexagon Lanai LoongArch MSP430 Mips NVPTX PowerPC RISCV Sparc SystemZ VE WebAssembly X86 XCore -ARC -CSKY -DirectX -M68k -SPIRV -Xtensa" PYTHON_SINGLE_TARGET="python3_12 -python3_10 -python3_11") 18.1.8(18/18.1)^t(11:19:05 04/07/24)(extra pie static-analyzer xml -debug -doc -ieee-long-double -test -verify-sig ABI_MIPS="-n32 -n64 -o32" ABI_S390="-32 -64" ABI_X86="64 -32 -x32" LLVM_TARGETS="AArch64 AMDGPU ARM AVR BPF Hexagon Lanai LoongArch MSP430 Mips NVPTX PowerPC RISCV Sparc SystemZ VE WebAssembly X86 XCore -ARC -CSKY -DirectX -M68k -SPIRV -Xtensa" PYTHON_SINGLE_TARGET="python3_12 -python3_10 -python3_11 -python3_13") Homepage: https://llvm.org/ Description: C language family frontend for LLVM [I] sys-devel/llvm Available versions: (15) 15.0.7-r3(15/15)^t (16) 16.0.6(16/16)^t (17) 17.0.6(17/17)^t (18) (~)18.1.8(18/18.1)^t (19) **19.0.0_pre20240623(19/19.0.0_pre20240623)^t **19.0.0.9999(19/19.0.0.9999)*l^t {+binutils-plugin (+)debug debuginfod doc exegesis libedit +libffi ncurses test verify-sig xar xml z3 zstd ABI_MIPS="n32 n64 o32" ABI_S390="32 64" ABI_X86="32 64 x32" LLVM_TARGETS="AArch64 AMDGPU ARC ARM AVR BPF CSKY DirectX Hexagon Lanai LoongArch M68k MSP430 Mips NVPTX PowerPC RISCV SPIRV Sparc SystemZ VE WebAssembly X86 XCore Xtensa"} Installed versions: 17.0.6(17/17)^t(04:30:54 22/06/24)(binutils-plugin libffi ncurses xml zstd -debug -debuginfod -doc -exegesis -libedit -test -verify-sig -xar -z3 ABI_MIPS="-n32 -n64 -o32" ABI_S390="-32 -64" ABI_X86="32 64 -x32" LLVM_TARGETS="AArch64 AMDGPU ARM AVR BPF Hexagon Lanai LoongArch MSP430 Mips NVPTX PowerPC RISCV Sparc SystemZ VE WebAssembly X86 XCore -ARC -CSKY -DirectX -M68k -SPIRV -Xtensa") 18.1.8(18/18.1)^t(10:57:42 04/07/24)(binutils-plugin libffi ncurses xml zstd -debug -debuginfod -doc -exegesis -libedit -test -verify-sig -z3 ABI_MIPS="-n32 -n64 -o32" ABI_S390="-32 -64" ABI_X86="32 64 -x32" LLVM_TARGETS="AArch64 AMDGPU ARM AVR BPF Hexagon Lanai LoongArch MSP430 Mips NVPTX PowerPC RISCV Sparc SystemZ VE WebAssembly X86 XCore -ARC -CSKY -DirectX -M68k -SPIRV -Xtensa") Homepage: https://llvm.org/ Description: Low Level Virtual Machine These are the packages that would be merged, in order: Calculating dependencies ... done! Dependency resolution took 2.84 s (backtrack: 0/20). [ebuild R ~] sys-devel/llvm-18.1.8:18/18.1::gentoo USE="binutils-plugin libffi ncurses xml zstd -debug -debuginfod -doc -exegesis -libedit -test -verify-sig -z3" ABI_X86="32 (64) (-x32)" LLVM_TARGETS="(AArch64) (AMDGPU) (ARM) (AVR) (BPF) (Hexagon) (Lanai) (LoongArch) (MSP430) (Mips) (NVPTX) (PowerPC) (RISCV) (Sparc) (SystemZ) (VE) (WebAssembly) (X86) (XCore) -ARC -CSKY -DirectX -M68k -SPIRV -Xtensa" 0 KiB Total: 1 package (1 reinstall), Size of downloads: 0 KiB ```
# portageq metadata / ebuild app-misc/foo-1.2.3 IUSE +llvm_slot_19 llvm_slot_18 Is this a stable system? Perhaps the problem you're hitting is that slot 18 is stable-masked.
Hi Michal, I've seen this on several global ~amd64 systems. While I have llvm 17 installed here I wanted to rule out an issue with it and tried to restrict the minimum slot in LLVM_COMPAT to 18. I expected the REQUIRED_USE to be automatically filled based on the eclass comments and was a bit surprised when it wasn't.
(In reply to Matt Jolly from comment #2) Your REQUIRED_USE is OK, I think (I got it wrong at first), the issue is that the IUSE default is wrong for you (or for some reason not working). mgorny's output shows the metadata looks right -- what does it show for you?
portageq metadata / ebuild app-misc/foo-1.2.3 IUSE +llvm_slot_19 llvm_slot_18
have you got LLVM_SLOT="18" in make.conf or package.use globally?
(also, it shouldn't be choosing 19 here, obviously, it should choose *18*)
No, nothing. Just grepped /etc/portage to be sure.
Ah, wait, so: * ^^ = exactly-one-of * 19 is set, so no +18 in there * nothing sets 18 So yeah, we're back to the issue being "it shouldn't choose +19, but instead +18, as 19 isn't released".
(In reply to Sam James from comment #8) > Ah, wait, so: > * ^^ = exactly-one-of > * 19 is set, so no +18 in there > * nothing sets 18 > > So yeah, we're back to the issue being "it shouldn't choose +19, but instead > +18, as 19 isn't released". Right, sorry, I've missed that in my output x_x.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=9f09b98ee6a5edce65e995c4701776e8e1c86d1d commit 9f09b98ee6a5edce65e995c4701776e8e1c86d1d Author: Michał Górny <mgorny@gentoo.org> AuthorDate: 2024-07-07 12:22:52 +0000 Commit: Michał Górny <mgorny@gentoo.org> CommitDate: 2024-07-23 14:07:52 +0000 llvm-r1.eclass: Do not default-enable unkeyworded slots Change the IUSE defaults logic to default-enable the *oldest* ~arch version rather than the newest one, when no stable slots are supported. Since we only except a single ~arch version to exist, this effectively prevents the eclass from default-enabling the unkeyworded snapshots. Closes: https://bugs.gentoo.org/935681 Signed-off-by: Michał Górny <mgorny@gentoo.org> Closes: https://github.com/gentoo/gentoo/pull/37473 Signed-off-by: Michał Górny <mgorny@gentoo.org> eclass/llvm-r1.eclass | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
