tools/include/nolibc/Makefile | 4 +- tools/include/nolibc/std.h | 15 +-- tools/include/nolibc/stdint.h | 84 ++++++++++++ tools/testing/selftests/nolibc/nolibc-test.c | 135 ++++++++++++------- 4 files changed, 173 insertions(+), 65 deletions(-) create mode 100644 tools/include/nolibc/stdint.h
Hi, This is version 3 of my patch to add stdint.h to nolibc. Previous versions of this patch are available here: * v2: https://lore.kernel.org/all/20230202201101.43160-1-v@vda.io/ * v1: https://lore.kernel.org/all/20230202160236.25342-1-v@vda.io/ This version should address most of the concerns raised by Willy on the previous version (missing LL, ULL suffixes, uintptr/size_max size). I tested it successfully on x86_64 and arm64 (on qemu). I have some trouble setting up my cross compilation chain for i386, arm and mips however. Vincent Dagonneau (4): tools/nolibc: Adding stdint.h tools/nolibc: Adding integer types and integer limit macros tools/nolibc: Enlarging column width of tests tools/nolibc: Adding tests for the integer limits in stdint.h tools/include/nolibc/Makefile | 4 +- tools/include/nolibc/std.h | 15 +-- tools/include/nolibc/stdint.h | 84 ++++++++++++ tools/testing/selftests/nolibc/nolibc-test.c | 135 ++++++++++++------- 4 files changed, 173 insertions(+), 65 deletions(-) create mode 100644 tools/include/nolibc/stdint.h -- 2.39.1
Hi Vincent,
On Sun, Feb 05, 2023 at 08:32:45PM -0500, Vincent Dagonneau wrote:
> Hi,
>
> This is version 3 of my patch to add stdint.h to nolibc. Previous
> versions of this patch are available here:
>
> * v2: https://lore.kernel.org/all/20230202201101.43160-1-v@vda.io/
> * v1: https://lore.kernel.org/all/20230202160236.25342-1-v@vda.io/
>
> This version should address most of the concerns raised by Willy on the
> previous version (missing LL, ULL suffixes, uintptr/size_max size).
OK thank you, that's much better already. I'm still having a few comments:
- please avoid the present participle in the subjects of your commits,
the imperative form is generally preferred (e.g: "add foo" instead of
"adding foo").
- do not forget your signed-off-by on patches.
- 3rd commit had an empty message. There's always something to say
about a change, at least why and what it provides.
- You still have this definition which is only valid for 64-bit,
you will need to move it to your ifdef __WORDSIZE block to adjust
it based on the word size:
#define SIZE_MAX UINT64_MAX
Maybe this one should also be part of the tests ?
> I tested it successfully on x86_64 and arm64 (on qemu). I have some
> trouble setting up my cross compilation chain for i386, arm and mips
> however.
What compilers do you use ? I'm using version 11.3 from these ones:
https://mirrors.edge.kernel.org/pub/tools/crosstool/
All supported archs work fine for me. Usually on x86_64 you can even
start the i386 binaries locally (if your kernel has COMPAT or COMPAT_32
built in, which is most often the case). And very commonly arm64 support
armv7 binaries as well.
Just let me know if I can help you set up your environment, it's always
a good long-term investment!
Regards,
Willy
On Mon, Feb 6, 2023, at 03:30, Willy Tarreau wrote: > Hi Vincent, > > On Sun, Feb 05, 2023 at 08:32:45PM -0500, Vincent Dagonneau wrote: >> Hi, >> >> This is version 3 of my patch to add stdint.h to nolibc. Previous >> versions of this patch are available here: >> >> * v2: https://lore.kernel.org/all/20230202201101.43160-1-v@vda.io/ >> * v1: https://lore.kernel.org/all/20230202160236.25342-1-v@vda.io/ >> >> This version should address most of the concerns raised by Willy on the >> previous version (missing LL, ULL suffixes, uintptr/size_max size). > > OK thank you, that's much better already. I'm still having a few comments: > - please avoid the present participle in the subjects of your commits, > the imperative form is generally preferred (e.g: "add foo" instead of > "adding foo"). > Noted. > - do not forget your signed-off-by on patches. > Again, sorry. I'm quite new to the send-email workflow. I just figured that I was probably supposed to use format-patch then send-email rather than just send-email... Makes way more sense though and now I can have a file with my cover letter and carry it over to the next version. > - 3rd commit had an empty message. There's always something to say > about a change, at least why and what it provides. > Noted. > - You still have this definition which is only valid for 64-bit, > you will need to move it to your ifdef __WORDSIZE block to adjust > it based on the word size: > > #define SIZE_MAX UINT64_MAX > > Maybe this one should also be part of the tests ? > Ok. >> I tested it successfully on x86_64 and arm64 (on qemu). I have some >> trouble setting up my cross compilation chain for i386, arm and mips >> however. > > What compilers do you use ? I'm using version 11.3 from these ones: > > https://mirrors.edge.kernel.org/pub/tools/crosstool/ > > All supported archs work fine for me. Usually on x86_64 you can even > start the i386 binaries locally (if your kernel has COMPAT or COMPAT_32 > built in, which is most often the case). And very commonly arm64 support > armv7 binaries as well. > > Just let me know if I can help you set up your environment, it's always > a good long-term investment! > Ah, foolishly I was relying on the packaged cross compiler for Arch. Compiled everything just fine with the packaged versions on Ubuntu (gcc-10 mips/arm/arm64/i386). I'll try with the toolchains @ https://mirrors.edge.kernel.org/pub/tools/crosstool/ as well. > Regards, > Willy Thanks, Vincent.
On Mon, Feb 06, 2023 at 05:19:18PM -0500, Vincent Dagonneau wrote: > > - do not forget your signed-off-by on patches. > > > > Again, sorry. I'm quite new to the send-email workflow. I just figured that I > was probably supposed to use format-patch then send-email rather than just > send-email... Makes way more sense though and now I can have a file with my > cover letter and carry it over to the next version. Don't be sorry. It can take a bit of time to get used to send-email (and can be scary at times). And yes, I, too, prefer to first produce the patches then send them. One advice is to produce them into a different directory each time, it will save you from accidently sending a mix of the previous version and the current one, and it will also help you double-check the differences between the patches to see check if you missed anything. > >> I tested it successfully on x86_64 and arm64 (on qemu). I have some > >> trouble setting up my cross compilation chain for i386, arm and mips > >> however. > > > > What compilers do you use ? I'm using version 11.3 from these ones: > > > > https://mirrors.edge.kernel.org/pub/tools/crosstool/ > > > > All supported archs work fine for me. Usually on x86_64 you can even > > start the i386 binaries locally (if your kernel has COMPAT or COMPAT_32 > > built in, which is most often the case). And very commonly arm64 support > > armv7 binaries as well. > > > > Just let me know if I can help you set up your environment, it's always > > a good long-term investment! > > > > Ah, foolishly I was relying on the packaged cross compiler for Arch. Compiled > everything just fine with the packaged versions on Ubuntu (gcc-10 > mips/arm/arm64/i386). I'll try with the toolchains @ > https://mirrors.edge.kernel.org/pub/tools/crosstool/ as well. OK! Cheers, Willy
© 2016 - 2025 Red Hat, Inc.