[PATCH v3] Changed ARM fenv support similar to X86_64

Eshan Dhawan eshandhawan51 at gmail.com
Sat Jul 18 15:39:28 UTC 2020

> On 18-Jul-2020, at 8:51 PM, Sebastian Huber <sebastian.huber at embedded-brains.de> wrote:
> On 18/07/2020 17:17, Eshan Dhawan wrote:
>>>>> On 18-Jul-2020, at 4:52 PM, Sebastian Huber <sebastian.huber at embedded-brains.de> wrote:
>>>> Hello Eshan,
>>>> you produce a lot of extra work for me. Your patch applies now. But it didn't build:
>>>> make[9]: *** No rule to make target 'fenv-vfp.c', needed by 'lib_a-fenv-vfp.o'.  Stop.
>>>> make[9]: *** Waiting for unfinished jobs....
>>>> make[9]: Leaving directory '/build/git-build/b-gcc-10.1.0-arm-rtems5/arm-rtems5/thumb/newlib/libm/machine/arm'
>> Fenv-vpf.c file was deleted as well as the declarations were deleted from makefile.am
>> So it shouldn’t be compiled
>> Did you try Remaking the autoconf files after applying the patch ???
>> This patch have the autoconf files
> I don't want to run autoconf myself to test your patches. Please include the build system changes in your patches.
Apologies, I will add the autoconf files from future patches.
>>> Why is x86_64mentioned in the commit message? On the mailing list it was discussed to use RISC-V as the model. Please improve the commit message in general. It should reflect all the problems discussed on the mailing list and how the commit addresses the issues.
>> The files follow the similar file structure as x86_64. So I added X86
> Which file did you use as a template?
The softfp Implementation is taken from riscv but has stub implementation 
>>> Attached is a patch which fixes a couple of obvious things. Please apply it, review it, test it, and make a v4 patch out of it with an improved commit message.
>> The error isn’t supposed to occur
>> I tested the code on machine
>>> <0001-arm-fenv.c-Fix-use-of-defines-and-format.patch>
>> I will apply the patch and test it
>> And resend it
>> Is this patch applied over the current patch
>> Or the newlib master?
> Over your v3 patch.
Ok, I will apply and test them.

More information about the devel mailing list