Hi Masahiro,
On 26 November 2014 at 00:45, Masahiro Yamada yamada.m@jp.panasonic.com wrote:
Hi Tom, Simon, other developers,
Since commit 0d296cc2d3b8 (Provide option to avoid defining a custom version of uintptr_t) and commit 4166ecb247a1 (Add some standard headers external code might need), I have been wondering if they were right decisions.
As arch/arm/include/asm/types.h of Linux Kernel says, using 'stdint.h' is not feasible.
-------------------------------------->8------------------------------------- /*
- The C99 types uintXX_t that are usually defined in 'stdint.h' are not as
- unambiguous on ARM as you would expect. For the types below, there is a
- difference on ARM between GCC built for bare metal ARM, GCC built for glibc
- and the kernel itself, which results in build errors if you try to build with
- -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h'
- in order to use NEON intrinsics)
- As the typedefs for these types in 'stdint.h' are based on builtin defines
- supplied by GCC, we can tweak these to align with the kernel's idea of those
- types, so 'linux/types.h' and 'stdint.h' can be safely included from the same
- source file (provided that -ffreestanding is used).
int32_t uint32_t uintptr_t- bare metal GCC long unsigned long unsigned int
- glibc GCC int unsigned int unsigned int
- kernel int unsigned int unsigned long
*/ --------------------------------------8<----------------------------------------
To me this doesn't matter. I feel that int32_t should probably be int on 32-bit ARM, but actually so long as it is consistent then it is fine if it is long. In fact so long as these types are defined consistently, everything works.
Actually, the kernel never includes <stdint.h> except host programs.
Commit 0d296cc2d3b8 introduced "USE_STDINT", but it causes type conflicts depending on which GCC is used.
With ARM bare metal GCC,
yamada@beagle:~/workspace/u-boot$ make omap3_beagle_defconfig all CROSS_COMPILE=arm-none-eabi- USE_STDINT=1 # # configuration written to .config # # # configuration written to spl/.config # scripts/kconfig/conf --silentoldconfig Kconfig scripts/kconfig/conf --silentoldconfig Kconfig CHK include/config.h GEN include/autoconf.mk GEN include/autoconf.mk.dep GEN spl/include/autoconf.mk CHK include/config/uboot.release CHK include/generated/version_autogenerated.h CHK include/generated/timestamp_autogenerated.h UPD include/generated/timestamp_autogenerated.h CC lib/asm-offsets.s In file included from /opt/arm-2011.03/bin/../lib/gcc/arm-none-eabi/4.5.2/include/stdint.h:5:0, from include/compiler.h:117, from include/image.h:19, from include/common.h:85, from lib/asm-offsets.c:15: /opt/arm-2011.03/bin/../lib/gcc/arm-none-eabi/4.5.2/include/stdint-gcc.h:40:24: error: conflicting types for 'int32_t' include/linux/types.h:99:17: note: previous declaration of 'int32_t' was here /opt/arm-2011.03/bin/../lib/gcc/arm-none-eabi/4.5.2/include/stdint-gcc.h:52:25: error: conflicting types for 'uint32_t' include/linux/types.h:105:17: note: previous declaration of 'uint32_t' was here make[2]: *** [lib/asm-offsets.s] Error 1 make[1]: *** [prepare0] Error 2 make: *** [__build_one_by_one] Error 2
While toolchain is this? I don't see this problem - it seems broken.
OK, we can fix "int32_t" and "uint32_t", but it still seems strange to see that "uint32_t" is defined as "unsigned long", whereas "u32" is defined as "unsigned int".
If so, must we fix "u32", "s32", ... all the fixed-width typedefs ?
I notice including <linux/types.h> in the U-Boot source tree and <stdint.h> provided by your compiler at the same time is a nightmare.
If we lean toward <stdint.h>, we must ban <linux/types.h>, but <stdint.h> is not available all the time. For example, kernel.org tool-chains do not provide <stdint.h>.
Maybe we can drastically re-write <linux/types.h> and friends to resolve the type conflicts, but I do not think we should not drift apart from the kernel because we have borrowed many source files from Linux.
So far I don't see a big problem. I feel that, were stdint.h available earlier, then types.h might not have been written and we would just use stdint.h. Presumably stdint.h has been created to fix these sorts of problems, and in fact for new projects, they would not define their own types.
IMO in time the kernel and U-Boot might move to stdint.h, and having it as an option at the moment helps us understand the issues. It does not break any builds.
I could be wrong though, time will tell. We should keep an eye on it.
Regards. Simon