Currently the test setup does not support running nolibc-test built with LLVM in qemu-system. Enable this.
FYI, sparc32 on LLVM seems to be broken at the moment. To me this looks like a LLVM regression, emitting invalid object code.
Signed-off-by: Thomas Weißschuh linux@weissschuh.net --- Thomas Weißschuh (3): selftests/nolibc: deduplicate invocations of toplevel Makefile selftests/nolibc: don't pass CC to toplevel Makefile selftests/nolibc: always compile the kernel with GCC
tools/testing/selftests/nolibc/Makefile.nolibc | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) --- base-commit: b9e50363178a40c76bebaf2f00faa2b0b6baf8d1 change-id: 20250719-nolibc-llvm-system-311762b62829
Best regards,
Various targets of the testsuite call back into the toplevel kernel Makefile. These calls use various parameters and are quite long.
Introduce a common variable to make future changes smaller and the lines shorter.
Signed-off-by: Thomas Weißschuh linux@weissschuh.net --- tools/testing/selftests/nolibc/Makefile.nolibc | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/tools/testing/selftests/nolibc/Makefile.nolibc b/tools/testing/selftests/nolibc/Makefile.nolibc index 0fb759ba992ee6b1693b88f1b2e77463afa9f38b..8e7a832ba3f6da3ca0d669b941acd1cf79813a8d 100644 --- a/tools/testing/selftests/nolibc/Makefile.nolibc +++ b/tools/testing/selftests/nolibc/Makefile.nolibc @@ -262,6 +262,9 @@ REPORT ?= awk '/[OK][\r]*$$/{p++} /[FAIL][\r]*$$/{if (!f) printf("\n"); f++ if (f || !p || !done) printf("failure\n"); else if (s) printf("warning\n"); else printf("success\n");; \ printf("\nSee all results in %s\n", ARGV[1]); }'
+# Execute the toplevel kernel Makefile +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) + help: @echo "Supported targets under selftests/nolibc:" @echo " all call the "run" target below" @@ -340,17 +343,17 @@ initramfs: nolibc-test $(Q)cp nolibc-test initramfs/init
defconfig: - $(Q)$(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) $(DEFCONFIG) + $(Q)$(KBUILD_MAKE) $(DEFCONFIG) $(Q)if [ -n "$(EXTRACONFIG)" ]; then \ $(srctree)/scripts/config --file $(objtree)/.config $(EXTRACONFIG); \ - $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) olddefconfig < /dev/null; \ + $(KBUILD_MAKE) olddefconfig < /dev/null; \ fi
kernel: - $(Q)$(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) $(IMAGE_NAME) < /dev/null + $(Q)$(KBUILD_MAKE) $(IMAGE_NAME) < /dev/null
kernel-standalone: initramfs - $(Q)$(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) $(IMAGE_NAME) CONFIG_INITRAMFS_SOURCE=$(CURDIR)/initramfs < /dev/null + $(Q)$(KBUILD_MAKE) $(IMAGE_NAME) CONFIG_INITRAMFS_SOURCE=$(CURDIR)/initramfs < /dev/null
# run the tests after building the kernel run: kernel initramfs.cpio
The toplevel Makefile is capable of calculating CC from CROSS_COMPILE and/or ARCH.
Stop passing the unnecessary variable.
Signed-off-by: Thomas Weißschuh linux@weissschuh.net --- tools/testing/selftests/nolibc/Makefile.nolibc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/nolibc/Makefile.nolibc b/tools/testing/selftests/nolibc/Makefile.nolibc index 8e7a832ba3f6da3ca0d669b941acd1cf79813a8d..6e6dddaa2e3ed56886cd76c423297648d13027c7 100644 --- a/tools/testing/selftests/nolibc/Makefile.nolibc +++ b/tools/testing/selftests/nolibc/Makefile.nolibc @@ -263,7 +263,7 @@ REPORT ?= awk '/[OK][\r]*$$/{p++} /[FAIL][\r]*$$/{if (!f) printf("\n"); f++ printf("\nSee all results in %s\n", ARGV[1]); }'
# Execute the toplevel kernel Makefile -KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE)
help: @echo "Supported targets under selftests/nolibc:"
Hi Thomas,
On Sat, Jul 19, 2025 at 05:38:28PM +0200, Thomas Weißschuh wrote:
The toplevel Makefile is capable of calculating CC from CROSS_COMPILE and/or ARCH.
Stop passing the unnecessary variable.
(...)
# Execute the toplevel kernel Makefile -KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE)
Here the goal was not to help the toplevel Makefile figure CC, but rather to permit the user to override it, and it's also listed in "make help", and even used in cc-option.
I understnad that you're trying to avoid forcing CC to clang when building, but in this case, what will CROSS_COMPILE contain ? My guess is that you intend to make CROSS_COMPILE point to the gcc-based toolchain, and have CC point to clang for userland only. Is this the case ? I think I'd be fine with this, but then we need to make it explicit in the help message and fix the current one, possibly just with this:
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)" + @echo " nolibc-test build the executable (uses $$CC)" @echo " libc-test build an executable using the compiler's default libc instead" @echo " run-user runs the executable under QEMU (uses $$XARCH, $$TEST)" @echo " initramfs.cpio prepare the initramfs archive with nolibc-test" @echo " initramfs prepare the initramfs tree with nolibc-test" @echo " defconfig create a fresh new default config (uses $$XARCH)" - @echo " kernel (re)build the kernel (uses $$XARCH)" + @echo " kernel (re)build the kernel (uses $$XARCH, $$CROSS_COMPILE)"
Thanks, Willy
Hi Willy,
On 2025-07-21 04:56:27+0200, Willy Tarreau wrote:
On Sat, Jul 19, 2025 at 05:38:28PM +0200, Thomas Weißschuh wrote:
The toplevel Makefile is capable of calculating CC from CROSS_COMPILE and/or ARCH.
Stop passing the unnecessary variable.
(...)
# Execute the toplevel kernel Makefile -KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE)
Here the goal was not to help the toplevel Makefile figure CC, but rather to permit the user to override it, and it's also listed in "make help", and even used in cc-option.
I understnad that you're trying to avoid forcing CC to clang when building, but in this case, what will CROSS_COMPILE contain ? My guess is that you intend to make CROSS_COMPILE point to the gcc-based toolchain, and have CC point to clang for userland only. Is this the case ?
Correct.
I think I'd be fine with this, but then we need to make it explicit in the help message and fix the current one, possibly just with this:
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC)"
I don't think this is correct. $CC itself depends on $CROSS_COMPILE through tools/scripts/Makefile.include.
@echo " libc-test build an executable using the compiler's default libc instead" @echo " run-user runs the executable under QEMU (uses $$XARCH, $$TEST)" @echo " initramfs.cpio prepare the initramfs archive with nolibc-test" @echo " initramfs prepare the initramfs tree with nolibc-test" @echo " defconfig create a fresh new default config (uses $$XARCH)"
- @echo " kernel (re)build the kernel (uses $$XARCH)"
- @echo " kernel (re)build the kernel (uses $$XARCH, $$CROSS_COMPILE)"
I'll fold this into the last commit of the series.
Thomas
Hi Thomas!
On Sun, Aug 17, 2025 at 10:30:52AM +0200, Thomas Weißschuh wrote:
Hi Willy,
On 2025-07-21 04:56:27+0200, Willy Tarreau wrote:
On Sat, Jul 19, 2025 at 05:38:28PM +0200, Thomas Weißschuh wrote:
The toplevel Makefile is capable of calculating CC from CROSS_COMPILE and/or ARCH.
Stop passing the unnecessary variable.
(...)
# Execute the toplevel kernel Makefile -KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CC=$(CC) CROSS_COMPILE=$(CROSS_COMPILE) +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE)
Here the goal was not to help the toplevel Makefile figure CC, but rather to permit the user to override it, and it's also listed in "make help", and even used in cc-option.
I understnad that you're trying to avoid forcing CC to clang when building, but in this case, what will CROSS_COMPILE contain ? My guess is that you intend to make CROSS_COMPILE point to the gcc-based toolchain, and have CC point to clang for userland only. Is this the case ?
Correct.
I think I'd be fine with this, but then we need to make it explicit in the help message and fix the current one, possibly just with this:
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC)"
I don't think this is correct. $CC itself depends on $CROSS_COMPILE through tools/scripts/Makefile.include.
I don't understand what you mean by "depends on" here. CC defaults to ${CROSS_COMPILE}gcc and may override it if set. So if one sets CC, CROSS_COMPILE will not be used for it. Or maybe we could change it to this to indicate a precedence if that's the idea you want to convey ?
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)" + @echo " nolibc-test build the executable (uses $$CC or $$CROSS_COMPILE)"
Willy
On 2025-08-17 11:39:05+0200, Willy Tarreau wrote:
On Sun, Aug 17, 2025 at 10:30:52AM +0200, Thomas Weißschuh wrote:
On 2025-07-21 04:56:27+0200, Willy Tarreau wrote:
On Sat, Jul 19, 2025 at 05:38:28PM +0200, Thomas Weißschuh wrote:
(...)
I think I'd be fine with this, but then we need to make it explicit in the help message and fix the current one, possibly just with this:
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC)"
I don't think this is correct. $CC itself depends on $CROSS_COMPILE through tools/scripts/Makefile.include.
I don't understand what you mean by "depends on" here. CC defaults to ${CROSS_COMPILE}gcc and may override it if set. So if one sets CC, CROSS_COMPILE will not be used for it. Or maybe we could change it to this to indicate a precedence if that's the idea you want to convey ?
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC or $$CROSS_COMPILE)"
Yes, that is the idea. I pushed a commit to this effect to nolibc/for-next.
Thomas
On Mon, Aug 18, 2025 at 04:07:21PM +0200, Thomas Weißschuh wrote:
On 2025-08-17 11:39:05+0200, Willy Tarreau wrote:
On Sun, Aug 17, 2025 at 10:30:52AM +0200, Thomas Weißschuh wrote:
On 2025-07-21 04:56:27+0200, Willy Tarreau wrote:
On Sat, Jul 19, 2025 at 05:38:28PM +0200, Thomas Weißschuh wrote:
(...)
I think I'd be fine with this, but then we need to make it explicit in the help message and fix the current one, possibly just with this:
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC)"
I don't think this is correct. $CC itself depends on $CROSS_COMPILE through tools/scripts/Makefile.include.
I don't understand what you mean by "depends on" here. CC defaults to ${CROSS_COMPILE}gcc and may override it if set. So if one sets CC, CROSS_COMPILE will not be used for it. Or maybe we could change it to this to indicate a precedence if that's the idea you want to convey ?
- @echo " nolibc-test build the executable (uses $$CC and $$CROSS_COMPILE)"
- @echo " nolibc-test build the executable (uses $$CC or $$CROSS_COMPILE)"
Yes, that is the idea. I pushed a commit to this effect to nolibc/for-next.
OK perfect, thank you! Willy
LLVM/clang can not build the kernel for all architectures supported by nolibc. The current setup uses the same compiler to build the kernel as is used for nolibc-test. This prevents using the full qemu-system tests for LLVM builds.
Instead always build the kernel with GCC. For the nolibc testsuite the kernel does not need to be built with LLVM.
Signed-off-by: Thomas Weißschuh linux@weissschuh.net --- tools/testing/selftests/nolibc/Makefile.nolibc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/nolibc/Makefile.nolibc b/tools/testing/selftests/nolibc/Makefile.nolibc index 6e6dddaa2e3ed56886cd76c423297648d13027c7..3d8e314c2433a2d032cf4f58ea1d6a2f430faaae 100644 --- a/tools/testing/selftests/nolibc/Makefile.nolibc +++ b/tools/testing/selftests/nolibc/Makefile.nolibc @@ -263,7 +263,7 @@ REPORT ?= awk '/[OK][\r]*$$/{p++} /[FAIL][\r]*$$/{if (!f) printf("\n"); f++ printf("\nSee all results in %s\n", ARGV[1]); }'
# Execute the toplevel kernel Makefile -KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE) +KBUILD_MAKE = $(MAKE) -C $(srctree) ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE) LLVM=
help: @echo "Supported targets under selftests/nolibc:"
On Sat, Jul 19, 2025 at 05:38:26PM +0200, Thomas Weißschuh wrote:
Currently the test setup does not support running nolibc-test built with LLVM in qemu-system. Enable this.
FYI, sparc32 on LLVM seems to be broken at the moment. To me this looks like a LLVM regression, emitting invalid object code.
Signed-off-by: Thomas Weißschuh linux@weissschuh.net
For the series, once the help message is adjusted:
Acked-by: Willy Tarreau w@1wt.eu
Thanks! Willy
linux-kselftest-mirror@lists.linaro.org
-
Thomas Weißschuh -
Willy Tarreau