Hi Tom, Quentin,
On Fri, 8 May 2026 at 11:23, Tom Rini <[email protected]> wrote:
>
> On Fri, May 08, 2026 at 01:55:41PM +0200, Quentin Schulz wrote:
> > Hi Tom,
> >
> > On 5/5/26 5:10 PM, Tom Rini wrote:
> > > On Tue, May 05, 2026 at 01:58:26PM +0200, Quentin Schulz wrote:
> > > > Hi Tom,
> > > >
> > > > On 5/4/26 8:04 PM, Tom Rini wrote:
> > > > > On Mon, May 04, 2026 at 03:20:27PM +0200, Quentin Schulz wrote:
> > > > [...]
> > > > > > The first two patches are for fixing pytest calling make directly
> > > > > > on a
> > > > > > "dirty" build directory as generated by buildman using the clang-20
> > > > > > toolchain, but pytest does not set HOSTCC and CC to clang-20,
> > > > > > resulting
> > > > > > in clang flags being used with gcc).
> > > > > >
> > > > > > One can simply reproduce it (after backporting up to patch
> > > > > > "scripts/Makefile.extrawarn: sync with 6.1") by doing so in the
> > > > > > container:
> > > > > >
> > > > > > make HOSTCC=clang-20 CC=clang-20 O=build/sandbox sandbox_defconfig
> > > > > > make HOSTCC=clang-20 CC=clang-20 O=build/sandbox -j`nproc
> > > > > > make O=build/sandbox u-boot-initial-env
> > > > > >
> > > > > > It'll print:
> > > > > > gcc: error: unrecognized command-line option '-mno-global-merge'
> > > > > > gcc: error: unrecognized command-line option
> > > > > > '-Wformat-invalid-specifier'
> > > > > > gcc: error: unrecognized command-line option '-Qunused-arguments'
> > > > > >
> > > > > > This is essentially what's run within test_env_initial_env_file in
> > > > > > test/py/tests/test_env.py.
> > > > > >
> > > > > > While I'm not unhappy with the patches to pytest, it feels to me we
> > > > > > may
> > > > > > be missing something else for the u-boot-initial-env make target to
> > > > > > update the kconfig symbol based on the toolchain being used when
> > > > > > called.
> > > > >
> > > > > Is there some other target in the Makefile that should catch this kind
> > > > > of change? It's not just that build target that needs to be remade,
> > > > > it's
> > > > > everything, yes?
> > > > >
> > > >
> > > > For the reproducer, one actually needs to do
> > > >
> > > > diff --git a/Makefile b/Makefile
> > > > index bd33437fdfd..9079131f81e 100644
> > > > --- a/Makefile
> > > > +++ b/Makefile
> > > > @@ -1007,15 +1007,15 @@ KBUILD_CFLAGS += $(call
> > > > cc-option,-Wno-format-nonliteral)
> > > > KBUILD_CFLAGS += $(call cc-disable-warning, address-of-packed-member)
> > > >
> > > > ifdef CONFIG_CC_IS_CLANG
> > > > -KBUILD_CPPFLAGS += $(call cc-option,-Qunused-arguments,)
> > > > -KBUILD_CFLAGS += $(call cc-disable-warning, format-invalid-specifier)
> > > > -KBUILD_CFLAGS += $(call cc-disable-warning, gnu)
> > > > +KBUILD_CPPFLAGS += -Qunused-arguments
> > > > +KBUILD_CFLAGS += -Wno-format-invalid-specifier
> > > > +KBUILD_CFLAGS += -Wno-gnu
> > > > # Quiet clang warning: comparison of unsigned expression < 0 is always
> > > > false
> > > > -KBUILD_CFLAGS += $(call cc-disable-warning, tautological-compare)
> > > > +KBUILD_CFLAGS += -Wno-tautological-compare
> > > > # CLANG uses a _MergedGlobals as optimization, but this breaks
> > > > modpost, as
> > > > the
> > > > # source of a reference will be _MergedGlobals and not on of the
> > > > whitelisted names.
> > > > # See modpost pattern 2
> > > > -KBUILD_CFLAGS += $(call cc-option, -mno-global-merge,)
> > > > +KBUILD_CFLAGS += -mno-global-merge
> > > > else
> > > >
> > > > # These warnings generated too much noise in a regular build.
> > > >
> > > > Which is what's being done since kernel commit a1494304346a ("kbuild:
> > > > add
> > > > all Clang-specific flags unconditionally"). I actually had this in an
> > > > earlier version of this patch series and forgot I removed it.
> > > >
> > > > This seems to apply to envtools and tools-only as well, though make
> > > > prompts
> > > > me for the desired state of symbols that are now reachable, e.g.
> > > > ARM64_CRC32. And if I rerun the same command, then it works. It seems
> > > > like
> > > > the config from the previous run is used? I don't know if it's related
> > > > but
> > > > those two commands call make directly ($(Q)$(MAKE) $(build)=<path>). Do
> > > > you
> > > > have something in mind maybe? Or some more commands to run?
> > >
> > > I think I partly misunderstood at first. It's not that we switch from
> > > one toolchain to another, it's that we can't (as easily) use gcc for the
> > > target and clang for host tools (and vice-versa) because of how flags
> > > are set now, yes? If so, how does this work in the kernel?
> > >
> >
> > I'm not sure what's happening so it's quite difficult to put into words.
> >
> > Apply the following to the master branch:
> >
> > diff --git a/Makefile b/Makefile
> > index 49d9740327e..9a2367f0a97 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -999,15 +999,15 @@ KBUILD_CFLAGS += $(call
> > cc-option,-Wno-format-nonliteral)
> > KBUILD_CFLAGS += $(call cc-disable-warning, address-of-packed-member)
> >
> > ifdef CONFIG_CC_IS_CLANG
> > -KBUILD_CPPFLAGS += $(call cc-option,-Qunused-arguments,)
> > -KBUILD_CFLAGS += $(call cc-disable-warning, format-invalid-specifier)
> > -KBUILD_CFLAGS += $(call cc-disable-warning, gnu)
> > +KBUILD_CPPFLAGS += -Qunused-arguments
> > +KBUILD_CFLAGS += -Wno-format-invalid-specifier
> > +KBUILD_CFLAGS += -Wno-gnu
> > # Quiet clang warning: comparison of unsigned expression < 0 is always
> > false
> > -KBUILD_CFLAGS += $(call cc-disable-warning, tautological-compare)
> > +KBUILD_CFLAGS += -Wno-tautological-compare
> > # CLANG uses a _MergedGlobals as optimization, but this breaks modpost, as
> > the
> > # source of a reference will be _MergedGlobals and not on of the
> > whitelisted names.
> > # See modpost pattern 2
> > -KBUILD_CFLAGS += $(call cc-option, -mno-global-merge,)
> > +KBUILD_CFLAGS += -mno-global-merge
> > KBUILD_CFLAGS += $(call cc-option, -fcatch-undefined-behavior)
> > KBUILD_CFLAGS += $(call cc-disable-warning, deprecated-non-prototype)
> > else
> >
> >
> > What I do:
> >
> > make CROSS_COMPILE="aarch64-linux-gnu-" O=build/puma HOSTCC=clang CC=clang
> > puma-rk3399_defconfig
> > make CROSS_COMPILE="aarch64-linux-gnu-" O=build/puma HOSTCC=clang CC=clang
> > V=1
> > make CROSS_COMPILE="aarch64-linux-gnu-" O=build/puma V=1
> >
> > That is, I configure the defconfig with clang selected as toolchain,
> > (successfully) build U-Boot with clang, then rebuild switching to gcc.
> >
> > What happens is that the last make command prompts me for Kconfig symbols
> > that are now reachable and thus kconfig doesn't know what to do with them.
> >
> > See the (NEW) entries below:
> >
> > scripts/kconfig/conf --syncconfig Kconfig
> > *
> > * Restart config...
> > *
> > *
> > * ARM architecture
> > *
> > Enable support for CRC32 instruction (ARM64_CRC32) [Y/n/?] (NEW)
> >
> > Timer clock frequency (COUNTER_FREQUENCY) [24000000] 24000000
> > Generate position-independent pre-relocation code (POSITION_INDEPENDENT)
> > [N/y/?] n
> > Specify the early stack pointer relative to the .bss section
> > (INIT_SP_RELATIVE) [N/y/?] n
> > Emit virtualizable instructions (KVM_VIRT_INS) [N/y/?] n
> > ARM GICV2 driver (DRIVER_GICV2) [N/y/?] n
> > ARM GICV3 ITS (GIC_V3_ITS) [N/y/?] n
> > ARM GICV3 GIC600 SUPPORT (GICV3_SUPPORT_GIC600) [N/y/?] n
> > Do not enable icache (SYS_ICACHE_OFF) [N/y/?] n
> > Do not enable icache in SPL (SPL_SYS_ICACHE_OFF) [N/y/?] n
> > Do not enable dcache (SYS_DCACHE_OFF) [N/y/?] n
> > Do not enable dcache in SPL (SPL_SYS_DCACHE_OFF) [N/y/?] n
> > CP15 based cache enabling support (SYS_ARM_CACHE_CP15) [N/y/?] n
> > MMU-based Paged Memory Management Support (SYS_ARM_MMU) [N/y/?] n
> > Use the ARM v7 PMSA Compliant MPU (SYS_ARM_MPU) [N/y/?] n
> > Select the ARM data write cache policy
> > > 1. Write-back (WB) (SYS_ARM_CACHE_WRITEBACK)
> > 2. Write-through (WT) (SYS_ARM_CACHE_WRITETHROUGH)
> > 3. Write allocation (WA) (SYS_ARM_CACHE_WRITEALLOC)
> > choice[1-3?]: 1
> > Enable ARCH_CPU_INIT (ARCH_CPU_INIT) [N/y/?] n
> > ARM Generic Timer support (SYS_ARCH_TIMER) [Y/n/?] y
> > Support for ARM SMC Calling Convention (SMCCC) (ARM_SMCCC) [Y/?] y
> > ARM PL310 L2 cache controller (SYS_L2_PL310) [N/y/?] n
> > ARM PL310 L2 cache controller in SPL (SPL_SYS_L2_PL310) [N/y/?] n
> > L2cache off (SYS_L2CACHE_OFF) [N/y/?] n
> > prepare BOOT0 header (ENABLE_ARM_SOC_BOOT0_HOOK) [Y/?] y
> > Use an assembly optimized implementation of memcpy (USE_ARCH_MEMCPY) [N/y/?]
> > (NEW)
> >
> > Use an assembly optimized implementation of memcpy for SPL
> > (SPL_USE_ARCH_MEMCPY) [N/y/?] n
> > Use an assembly optimized implementation of memcpy for TPL
> > (TPL_USE_ARCH_MEMCPY) [N/y/?] n
> > Use an assembly optimized implementation of memset (USE_ARCH_MEMSET) [N/y/?]
> > (NEW)
> >
> > Use an assembly optimized implementation of memset for SPL
> > (SPL_USE_ARCH_MEMSET) [N/y/?] n
> > Use an assembly optimized implementation of memset for TPL
> > (TPL_USE_ARCH_MEMSET) [N/y/?] n
> > ARM64 system support AArch32 execution state (ARM64_SUPPORT_AARCH32) [Y/n/?]
> > y
> > Target select
> > 1. Airoha SoCs (ARCH_AIROHA)
> > 2. Atmel AT91 (ARCH_AT91)
> > 3. TI DaVinci (ARCH_DAVINCI)
> > 4. Hisilicon HiSTB SoCs (ARCH_HISTB)
> > 5. Marvell Kirkwood (ARCH_KIRKWOOD)
> > 6. Marvell MVEBU family (Armada XP/375/38x/3700/7K/8K) (ARCH_MVEBU)
> > 7. Broadcom BCM283X family (ARCH_BCM283X)
> > 8. Broadcom BCM7XXX family (ARCH_BCMSTB)
> > 9. Broadcom broadband chip family (ARCH_BCMBCA)
> > 10. Support vexpress_ca9x4 (TARGET_VEXPRESS_CA9X4)
> > 11. Support Broadcom Northstar (TARGET_BCMNS)
> > 12. Support Broadcom NS3 (TARGET_BCMNS3)
> > 13. Samsung EXYNOS (ARCH_EXYNOS)
> > 14. Samsung S5PC1XX (ARCH_S5PC1XX)
> > 15. Calxeda Highbank (ARCH_HIGHBANK)
> > 16. ARM Ltd. Integrator family (ARCH_INTEGRATOR)
> > 17. Qualcomm IPQ40xx SoCs (ARCH_IPQ40XX)
> > 18. TI Keystone (ARCH_KEYSTONE)
> > 19. Texas Instruments' K3 Architecture (ARCH_K3)
> > 20. TI OMAP2+ (ARCH_OMAP2PLUS)
> > 21. Amlogic Meson (ARCH_MESON)
> > 22. MediaTek SoCs (ARCH_MEDIATEK)
> > 23. Marvell MMP (ARCH_MMP)
> > 24. NXP LPC32xx platform (ARCH_LPC32XX)
> > 25. NXP i.MX8 platform (ARCH_IMX8)
> > 26. NXP i.MX8M platform (ARCH_IMX8M)
> > 27. NXP i.MX8ULP platform (ARCH_IMX8ULP)
> > 28. NXP i.MX9 platform (ARCH_IMX9)
> > 29. NXP i.MXRT platform (ARCH_IMXRT)
> > 30. NXP i.MX23 family (ARCH_MX23)
> > 31. NXP i.MX28 family (ARCH_MX28)
> > 32. NXP MX7ULP (ARCH_MX7ULP)
> > 33. Freescale MX7 (ARCH_MX7)
> > 34. Freescale MX6 (ARCH_MX6)
> > 35. Freescale MX5 (ARCH_MX5)
> > 36. Nexell S5P4418/S5P6818 SoC (ARCH_NEXELL)
> > 37. Support Nuvoton SoCs (ARCH_NPCM)
> > 38. Apple SoCs (ARCH_APPLE)
> > 39. Actions Semi OWL SoCs (ARCH_OWL)
> > 40. QEMU Virtual Platform (ARCH_QEMU)
> > 41. Renesas ARM SoCs (ARCH_RENESAS)
> > 42. Qualcomm Snapdragon SoCs (ARCH_SNAPDRAGON)
> > 43. Altera SOCFPGA family (ARCH_SOCFPGA)
> > 44. Support sunxi (Allwinner) SoCs (ARCH_SUNXI)
> > 45. ST-Ericsson U8500 Series (ARCH_U8500)
> > 46. Support Xilinx Versal Platform (ARCH_VERSAL)
> > 47. Support AMD Versal Gen 2 Platform (ARCH_VERSAL2)
> > 48. Support Xilinx Versal NET Platform (ARCH_VERSAL_NET)
> > 49. Freescale Vybrid (ARCH_VF610)
> > 50. Xilinx Zynq based platform (ARCH_ZYNQ)
> > 51. Xilinx ZynqMP R5 based platform (ARCH_ZYNQMP_R5)
> > 52. Xilinx ZynqMP based platform (ARCH_ZYNQMP)
> > 53. NVIDIA Tegra (ARCH_TEGRA)
> > 54. Support ARMv8 Arm Ltd. VExpress based boards and models
> > (ARCH_VEXPRESS64)
> > 55. Support Corstone1000 Platform (TARGET_CORSTONE1000)
> > 56. Support Total Compute Platform (TARGET_TOTAL_COMPUTE)
> > 57. Support ls2080a_emu (TARGET_LS2080A_EMU)
> > 58. Support ls1088aqds (TARGET_LS1088AQDS)
> > 59. Support ls2080aqds (TARGET_LS2080AQDS)
> > 60. Support ls2080ardb (TARGET_LS2080ARDB)
> > 61. Support ls2081ardb (TARGET_LS2081ARDB)
> > 62. Support lx2160ardb (TARGET_LX2160ARDB)
> > 63. Support lx2160aqds (TARGET_LX2160AQDS)
> > 64. Support lx2162aqds (TARGET_LX2162AQDS)
> > 65. Support HiKey 96boards Consumer Edition Platform (TARGET_HIKEY)
> > 66. Support HiKey960 96boards Consumer Edition Platform (TARGET_HIKEY960)
> > 67. Support Poplar 96boards Enterprise Edition Platform (TARGET_POPLAR)
> > 68. Support ls1012aqds (TARGET_LS1012AQDS)
> > 69. Support ls1012ardb (TARGET_LS1012ARDB)
> > 70. Support ls1012a2g5rdb (TARGET_LS1012A2G5RDB)
> > 71. Support ls1012afrwy (TARGET_LS1012AFRWY)
> > 72. Support ls1012afrdm (TARGET_LS1012AFRDM)
> > 73. Support ls1028aqds (TARGET_LS1028AQDS)
> > 74. Support ls1028ardb (TARGET_LS1028ARDB)
> > 75. Support ls1088ardb (TARGET_LS1088ARDB)
> > 76. Support ls1021aqds (TARGET_LS1021AQDS)
> > 77. Support ls1021atwr (TARGET_LS1021ATWR)
> > 78. Support Hitachi-Powergrids SELI8 service unit card
> > (TARGET_PG_WCOM_SELI8)
> > 79. Support Hitachi-Powergrids EXPU1 service unit card
> > (TARGET_PG_WCOM_EXPU1)
> > 80. Support ls1021atsn (TARGET_LS1021ATSN)
> > 81. Support ls1021aiot (TARGET_LS1021AIOT)
> > 82. Support ls1043aqds (TARGET_LS1043AQDS)
> > 83. Support ls1043ardb (TARGET_LS1043ARDB)
> > 84. Support ls1046aqds (TARGET_LS1046AQDS)
> > 85. Support ls1046ardb (TARGET_LS1046ARDB)
> > 86. Support ls1046afrwy (TARGET_LS1046AFRWY)
> > 87. Analog Devices SC5XX-processor family (ARCH_SC5XX)
> > 88. Support sl28 (TARGET_SL28)
> > 89. Support ten64 (TARGET_TEN64)
> > 90. Socionext UniPhier SoCs (ARCH_UNIPHIER)
> > 91. Socionext SynQuacer SoCs (ARCH_SYNQUACER)
> > 92. Support STMicroelectronics STM32 MCU with cortex M (ARCH_STM32)
> > 93. Support STMicroelectronics SoCs (ARCH_STI)
> > 94. Support STMicroelectronics STM32MP Socs with cortex A (ARCH_STM32MP)
> > > 95. Support Rockchip SoCs (ARCH_ROCKCHIP)
> > 96. Support OcteonTX SoCs (ARCH_OCTEONTX)
> > 97. Support OcteonTX2 SoCs (ARCH_OCTEONTX2)
> > 98. Support ThunderX 88xx (TARGET_THUNDERX_88XX)
> > 99. Support Aspeed SoCs (ARCH_ASPEED)
> > 100. Support Phytium Durian Platform (TARGET_DURIAN)
> > 101. Support Phytium Pomelo Platform (TARGET_POMELO)
> > 102. Support Phytium PE2201 Platform (TARGET_PE2201)
> > 103. Support Cortina Presidio ASIC Platform (TARGET_PRESIDIO_ASIC)
> > 104. Xen guest ARM64 (TARGET_XENGUEST_ARM64)
> > 105. Support HPE GXP SoCs (ARCH_GXP)
> > choice[1-105?]: 95
> > Support Rockchip PX30 (ROCKCHIP_PX30) [N/y/?] n
> > Support Rockchip RK3036 (ROCKCHIP_RK3036) [N/y/?] n
> > Support Rockchip RK3066 (ROCKCHIP_RK3066) [N/y/?] n
> > Support Rockchip RK3128 (ROCKCHIP_RK3128) [N/y/?] n
> > Support Rockchip RK3188 (ROCKCHIP_RK3188) [N/y/?] n
> > Support Rockchip RK3228/RK3229 (ROCKCHIP_RK322X) [N/y/?] n
> > Support Rockchip RK3288 (ROCKCHIP_RK3288) [N/y/?] n
> > Support Rockchip RK3308 (ROCKCHIP_RK3308) [N/y/?] n
> > Support Rockchip RK3328 (ROCKCHIP_RK3328) [N/y/?] n
> > Support Rockchip RK3368 (ROCKCHIP_RK3368) [N/y/?] n
> > Support Rockchip RK3399 (ROCKCHIP_RK3399) [Y/n/?] y
> > Support Rockchip RK3506 (ROCKCHIP_RK3506) [N/y/?] n
> > Support Rockchip RK3528 (ROCKCHIP_RK3528) [N/y/?] n
> > Support Rockchip RK3568 (ROCKCHIP_RK3568) [N/y/?] n
> > Support Rockchip RK3576 (ROCKCHIP_RK3576) [N/y/?] n
> > Support Rockchip RK3588 (ROCKCHIP_RK3588) [N/y/?] n
> > Support Rockchip RV1108 (ROCKCHIP_RV1108) [N/y/?] n
> > Support Rockchip RV1126 (ROCKCHIP_RV1126) [N/y/?] n
> > Route uart output to usb pins (ROCKCHIP_USB_UART) [N/y/?] n
> > SPL returns to bootrom (SPL_ROCKCHIP_BACK_TO_BROM) [N/y/?] n
> > TPL returns to bootrom (TPL_ROCKCHIP_BACK_TO_BROM) [Y/n/?] y
> > Rockchip common board file (ROCKCHIP_COMMON_BOARD) [Y/n/?] y
> > Rockchip SPL common board file (SPL_ROCKCHIP_COMMON_BOARD) [Y/n/?] y
> > Rockchip TPL common board file (TPL_ROCKCHIP_COMMON_BOARD) [Y/n/?] y
> > Use external TPL binary (ROCKCHIP_EXTERNAL_TPL) [N/y/?] n
> > Rockchip boot mode flag register address (ROCKCHIP_BOOT_MODE_REG) [0x0] 0x0
> > Disable device boot on power plug-in
> > (ROCKCHIP_RK8XX_DISABLE_BOOT_ON_POWERON) [N/y/?] n
> > Rockchip STIMER support (ROCKCHIP_STIMER) [Y/n/?] y
> > Size of IRAM reserved in SPL (ROCKCHIP_SPL_RESERVE_IRAM) [0x0] 0x0
> > SPL requires early-return (for RK3188-style BROM) to BROM
> > (SPL_ROCKCHIP_EARLYRETURN_TO_BROM) [N/y/?] n
> > Disable force_jtag feature (ROCKCHIP_DISABLE_FORCE_JTAG) [Y/n/?] y
> > TPL requires early-return (for RK3188-style BROM) to BROM
> > (TPL_ROCKCHIP_EARLYRETURN_TO_BROM) [N/y/?] n
> > Build a SPI image for rockchip (ROCKCHIP_SPI_IMAGE) [Y/n/?] y
> > Build a maskrom mode image for Rockchip (ROCKCHIP_MASKROM_IMAGE) [N/y/?] n
> > RK3399 board select
> > 1. Asus Flip C101PA Chromebook (RK3399) (TARGET_CHROMEBOOK_BOB)
> > 2. Samsung Chromebook Plus (RK3399) (TARGET_CHROMEBOOK_KEVIN)
> > 3. RK3399 evaluation board (TARGET_EVB_RK3399)
> > 4. Pinebook Pro (TARGET_PINEBOOK_PRO_RK3399)
> > 5. PinePhone Pro (TARGET_PINEPHONE_PRO_RK3399)
> > > 6. Theobroma Systems RK3399-Q7 (Puma) (TARGET_PUMA_RK3399)
> > 7. Vamrs Limited Rock960 board family (TARGET_ROCK960_RK3399)
> > 8. Radxa ROCK Pi 4 board (TARGET_ROCKPI4_RK3399)
> > 9. Pine64 Rockpro64 board (TARGET_ROCKPRO64_RK3399)
> > 10. Firefly ROC-RK3399-PC board (TARGET_ROC_PC_RK3399)
> > choice[1-10?]: 6
> > Theobroma Systems RK3399-Q7 DDR Option
> > > 1. DDR3-1333MHz (TARGET_PUMA_RK3399_RAM_DDR3_1333)
> > 2. DDR3-1600MHz (TARGET_PUMA_RK3399_RAM_DDR3_1600)
> > 3. DDR3-1866MHz (TARGET_PUMA_RK3399_RAM_DDR3_1866)
> > choice[1-3?]: 1
> > Force cache maintenance to be exclusively by VA (CMO_BY_VA_ONLY) [N/y/?] n
> > Fix broken ARMv8 generic timer (ARMV8_CNTFRQ_BROKEN) [N/y/?] n
> > Install crash dump exception vectors (ARMV8_SPL_EXCEPTION_VECTORS) [N/y/?] n
> > Enable multiple CPUs to enter into U-Boot (ARMV8_MULTIENTRY) [N/y/?] n
> > Enable data coherency with other cores in cluster (ARMV8_SET_SMPEN) [Y/n/?]
> > y
> > Enable switching to running in EL1 (ARMV8_SWITCH_TO_EL1) [N/y/?] n
> > Use PSCI for reset and shutdown (PSCI_RESET) [Y/n/?] y
> > Enable PSCI support (ARMV8_PSCI) [N/y/?] n
> > External aborts and SError interrupt exception are taken in EL3
> > (ARMV8_EA_EL3_FIRST) [N/y/?] n
> > Use the event stream for udelay (ARMV8_UDELAY_EVENT_STREAM) [N/y/?] n
> > *
> > * General setup
> > *
> > Compile also drivers which will not load (COMPILE_TEST) [N/y/?] n
> > Compile U-Boot with warnings as errors (WERROR) [N/y/?] n
> > Local version - append to U-Boot release (LOCALVERSION) []
> > Automatically append version information to the version string
> > (LOCALVERSION_AUTO) [Y/n/?] y
> > Optimization level
> > > 1. Optimize for size (CC_OPTIMIZE_FOR_SIZE)
> > 2. Optimize for speed (CC_OPTIMIZE_FOR_SPEED)
> > 3. Optimize for debugging (CC_OPTIMIZE_FOR_DEBUG)
> > choice[1-3?]: 1
> > Allow compiler to uninline functions marked 'inline' in full U-Boot
> > (OPTIMIZE_INLINING) [N/y/?] n
> > Allow compiler to uninline functions marked 'inline' in SPL
> > (SPL_OPTIMIZE_INLINING) [N/y/?] n
> > Allow compiler to uninline functions marked 'inline' in TPL
> > (TPL_OPTIMIZE_INLINING) [N/y/?] n
> > Enable Link Time Optimizations (LTO) [N/y/?] n
> > Select U-Boot be run as a bootloader for XEN Virtual Machine (XEN) [N/y/?] n
> > Add arch, board, vendor and soc variables to default environment
> > (ENV_VARS_UBOOT_CONFIG) [Y/?] y
> > Number of DRAM banks (NR_DRAM_BANKS) [1] 1
> > Enable kernel command line setup (SYS_BOOT_GET_CMDLINE) [N/y/?] n
> > Enable kernel board information setup (SYS_BOOT_GET_KBD) [N/y/?] n
> > Use a custom location for the initial stack pointer address
> > (HAS_CUSTOM_SYS_INIT_SP_ADDR) [Y/?] y
> > Static location for the initial stack pointer (CUSTOM_SYS_INIT_SP_ADDR)
> > [0x03f00000] 0x03f00000
> > Enable malloc() pool before relocation (SYS_MALLOC_F) [Y/?] y
> > Size of malloc() pool before relocation (SYS_MALLOC_F_LEN) [0x10000]
> > 0x10000
> > Define memory for Dynamic allocation (SYS_MALLOC_LEN) [0x2000000] 0x2000000
> > Enable malloc() pool in SPL (SPL_SYS_MALLOC_F) [Y/n/?] y
> > Size of malloc() pool in SPL (SPL_SYS_MALLOC_F_LEN) [0x8000] 0x8000
> > Enable malloc() pool in TPL (TPL_SYS_MALLOC_F) [Y/n/?] y
> > Size of malloc() pool in TPL (TPL_SYS_MALLOC_F_LEN) [0x4000] 0x4000
> > Inform valgrind about memory allocations (VALGRIND) [N/y/?] n
> > 64bit physical address support (PHYS_64BIT) [Y/?] y
> > 64bit fdt address support (FDT_64BIT) [Y/?] y
> > SPL image used in the combined SPL+U-Boot image (SPL_IMAGE)
> > [spl/u-boot-spl.bin] spl/u-boot-spl.bin
> > Recreate an ELF image from raw U-Boot binary (REMAKE_ELF) [N/y/?] n
> > Build target special images (BUILD_TARGET) []
> > Define a maximum size for the U-Boot image (HAS_BOARD_SIZE_LIMIT) [N/y/?] n
> > Use a custom location for the U-Boot linker script (SYS_CUSTOM_LDSCRIPT)
> > [N/y/?] n
> > Address in memory to use by default (SYS_LOAD_ADDR) [0x800800] 0x800800
> > Define max stack size that can be used by U-Boot (STACK_SIZE) [0x1000000]
> > 0x1000000
> > Exclude some memory from U-Boot / OS information (SYS_MEM_TOP_HIDE) [0x0]
> > 0x0
> > Maximum size in bytes reserved for U-Boot in memory (SYS_MONITOR_LEN) [0] 0
> > Support for multiprocessor (MP) [N/y/?] n
> > Text Base (TEXT_BASE) [0x00800000] 0x00800000
> > Use custom U-Boot Start (HAVE_SYS_UBOOT_START) [N/y/?] n
> > Determine CPU clock frequency at run-time (DYNAMIC_SYS_CLK_FREQ) [N/y/?] n
> > CPU clock frequency (SYS_CLK_FREQ) [0] 0
> > Enable U-Boot API (API) [N/y/?] (NEW)
> >
> > Then it continues with the build, but using the flags for clang while using
> > GCC, see:
> >
> > make -f /home/qschulz/work/upstream/u-boot/scripts/Makefile.autoconf || \
> > { rm -f include/config/auto.conf; false; }
> > mkdir -p include/asm
> > if [ -d /home/qschulz/work/upstream/u-boot/arch/arm/mach-rk3399/include/mach
> > ]; then \
> > dest=arch/arm/mach-rk3399/include/mach; \
> > else \
> > dest=arch/arm/include/asm/arch-rk3399; \
> > fi; \
> > ln -fsn /home/qschulz/work/upstream/u-boot/$dest include/asm/arch
> > set -e; mkdir -p include/; trap "rm -f include/.config.h.tmp" EXIT; { (echo
> > "/* Automatically generated - do not edit */"; echo \#define CFG_BOARDDIR
> > board/theobroma-systems/puma_rk3399; echo \#include
> > \<configs/"puma_rk3399".h\> ; echo \#include \<asm/config.h\>; echo
> > \#include \<linux/kconfig.h\>; echo \#include \<config_fallbacks.h\>;); } >
> > include/.config.h.tmp; if [ ! -r include/config.h ] || ! cmp -s
> > include/config.h include/.config.h.tmp; then : ' UPD include/config.h'; mv
> > -f include/.config.h.tmp include/config.h; fi
> > aarch64-linux-gnu-gcc -E -Wall -Werror=strict-prototypes -Wno-trigraphs
> > -fno-strict-aliasing -fno-common -fshort-wchar -fno-PIE
> > -Werror=implicit-function-declaration -Werror=implicit-int
> > -Wno-format-security -std=gnu11 -ffreestanding -fno-builtin
> > -fno-delete-null-pointer-checks -Wno-frame-address -Wno-format-truncation
> > -Wno-format-overflow -Wno-int-in-bool-context -Wno-address-of-packed-member
> > -Os -fno-stack-protector -fno-delete-null-pointer-checks
> > -Wno-zero-length-bounds -Wno-array-bounds -Wno-stringop-overflow
> > -Wno-maybe-uninitialized
> > -fmacro-prefix-map=/home/qschulz/work/upstream/u-boot/= -gdwarf-4
> > -fstack-usage -Wno-format-nonliteral -Wno-address-of-packed-member
> > -Wno-format-invalid-specifier -Wno-gnu -Wno-tautological-compare
> > -mno-global-merge -Wno-pointer-sign -D__KERNEL__ -D__UBOOT__
> > -Qunused-arguments -D__ARM__ -ffunction-sections -fdata-sections
> > -fno-common -ffixed-x18 -pipe -Iinclude
> > -I/home/qschulz/work/upstream/u-boot/include
> > -I/home/qschulz/work/upstream/u-boot/arch/arm/include -include
> > /home/qschulz/work/upstream/u-boot/include/linux/kconfig.h -nostdinc
> > -isystem /usr/lib/gcc/aarch64-linux-gnu/15/include -DDO_DEPS_ONLY -dM
> > include/config.h > u-boot.cfg.tmp && { grep 'define CONFIG_' u-boot.cfg.tmp
> > | sed '/define CONFIG_IS_ENABLED(/d;/define CONFIG_IF_ENABLED_INT(/d;/define
> > CONFIG_VAL(/d;' > u-boot.cfg; rm u-boot.cfg.tmp; } || { rm u-boot.cfg.tmp;
> > false; }
> > aarch64-linux-gnu-gcc: error: unrecognized command-line option
> > ‘-mno-global-merge’; did you mean ‘-Wno-global-module’?
> > aarch64-linux-gnu-gcc: error: unrecognized command-line option
> > ‘-Qunused-arguments’
> > make[2]: ***
> > [/home/qschulz/work/upstream/u-boot/scripts/Makefile.autoconf:84:
> > u-boot.cfg] Error 1
> > make[1]: *** No rule to make target 'include/config/auto.conf', needed by
> > 'include/config/uboot.release'. Stop.
> > make: *** [Makefile:189: __sub-make] Error 2
> >
> > If I do nothing but call
> > make CROSS_COMPILE="aarch64-linux-gnu-" O=build/puma V=1
> > again, it'll work just fine and compile with GCC flags with the GCC
> > toolchain (well, I don't get an error anymore, so I'm assumign that what's
> > happening :) ). So it seems to me that the symbols as used in Makefile
> > depend on the previous make command generating the .config and syncconfig is
> > running "too late".
> >
> > Contrary to what I initially thought, it's not actually related to U-Boot
> > host tools.
> >
> > I'm not attempting to mix GCC and clang for host/target building, and I
> > don't know if this would work in any case. I don't know if this is something
> > we want to support anyway. As for what the Linux kernel does, they
> > document[1] LLVM/clang support by telling the user to essentially use LLVM=1
> > which they claim is an alias to:
> > CC=clang LD=ld.lld AR=llvm-ar NM=llvm-nm STRIP=llvm-strip \
> > OBJCOPY=llvm-objcopy OBJDUMP=llvm-objdump READELF=llvm-readelf \
> > HOSTCC=clang HOSTCXX=clang++ HOSTAR=llvm-ar HOSTLD=ld.lld
> >
> > They don't really hint at CC and HOSTCC being different as being an option.
> > Maybe it works, maybe it doesn't?
> >
> > I did test in the Linux kernel whether they had the same issue but it
> > doesn't seem like it is the case.
> >
> > [1] https://www.kernel.org/doc/html/latest/kbuild/llvm.html
>
> Ah, thanks. I'm tempted to say this is something that'll just not work
> until we can also switch to following the kernel way of invoking clang?
> But maybe it's a problem missing dependencies somewhere (not leading to
> some cmd file being recreated in time), rather than more missing
> changes.
The root cause is likely a stale KBUILD_CFLAGS leaking into
Makefile.autoconf sub-make.
The U-Boot-specific recipe at Makefile:813 runs two steps when
'auto.conf' is stale:
1. 'make syncconfig' - regenerates 'auto.conf' (e.g. flips
CC_IS_CLANG=y to CC_IS_GCC=y after a toolchain change).
2. 'make -f scripts/Makefile.autoconf' - preprocesses
'include/config.h' to produce u-boot.cfg/autoconf.mk
KBUILD_CFLAGS and KBUILD_CPPFLAGS are exported by the parent (line
576) and consumed by scripts/Makefile.autoconf:45:
c_flags := $(KBUILD_CFLAGS) $(KBUILD_CPPFLAGS) ...
The parent computes those at parse time from the stale 'auto.conf', so
step 2's sub-make inherits clang-only flags ('-mno-global-merge',
'-Qunused-arguments', ...) and feeds them to gcc, fails. The recipe
deletes 'auto.conf' and aborts, so GNU make's makefile-restart never
runs. A second 'make' succeeds because 'auto.conf' is already fresh on
disk before parsing.
Linux doesn't hit this: its auto.conf rule only runs 'syncconfig' and
lets make re-exec. Basically Makefile.autoconf is a old bridge for the
include/configs -> Kconfig migration.
We should stop leaking parent parse-time CFLAGS into
'Makefile.autoconf'. That helper only does 'cpp -dM' on 'config.h' and
needs UBOOTINCLUDE, PLATFORM_CPPFLAGS, '-D__KERNEL__', etc. - not
warning/codegen flags. Either 'unexport KBUILD_CFLAGS KBUILD_CPPFLAGS'
for that sub-make, or reset them inside 'Makefile.autoconf' after
'include config.mk' and rebuild only the cpp pieces it actually needs.
This also explains Quentin's hunch about 'u-boot-initial-env': any
target that triggers the auto.conf recipe runs 'Makefile.autoconf'
with the parent's stale env, regardless of which target was asked for.
Something like this before the CC = in Makefile.autoconf:
# Discard KBUILD_CFLAGS/KBUILD_CPPFLAGS inherited from the parent make.
# They were computed at the parent's parse time from a possibly-stale
# auto.conf (e.g. before syncconfig detected a toolchain change), so
# CC_IS_CLANG-guarded flags can leak across a clang->gcc switch and
# break the cpp -dM below. Warning/codegen flags do not affect cpp
# -dM output, so only the base preprocessor defines and CLANG_FLAGS
# (--target=, --gcc-toolchain=, ...) need to be preserved here.
KBUILD_CFLAGS := $(CLANG_FLAGS)
KBUILD_CPPFLAGS := -D__KERNEL__ -D__UBOOT__ $(CLANG_FLAGS)
CLANG_FLAGS must stay otherwise cross-compile clang builds break:
without '--target=aarch64-linux-gnu' clang falls back to the host
triple and rejects arch-specific flags from PLATFORM_CPPFLAGS like
'-ffixed-x18' (arch/arm/config.mk).
Regards,
Simon
