[R-SIG-Win] GCC 5.3.0 - sd(NA) = NaN?

[Avraham Adler]

Hi, Jeroen.

Thanks for the link. Has anyone posted the possible incorrect
implementation at mingw64 for their review?

Thanks,

Avi

On Fri, Mar 18, 2016 at 5:32 AM, Jeroen Ooms <jeroenooms at gmail.com> wrote:
> I found something similar a while back:
> http://r.789695.n4.nabble.com/Optimization-bug-when-byte-compiling-with-gcc-5-2-0-on-windows-td4712220.html
>
> IIRC the issue was diagnosed by Luke that there are a few workarounds
> that currently use: #if __GNUC__ <= 4. However when changing that to
> 5, all kind of other problems appeared.
>
>
>
> On Thu, Mar 17, 2016 at 6:34 PM, Avraham Adler <avraham.adler at gmail.com> wrote:
>> Hello.
>>
>> As an experiment, I tried compiling R with the mingw64 5.3.0 (posix-seh)
>> build. R compiles smoothly, but make check fails pretty early as sd(NA)
>> returns NaN instead of NA, which is a fatal error in the tests.I have no
>> idea what may cause this, but thought it may be of interest to this group.
>> I'd appreciate it if someone who understands why that happens would post a
>> brief explanation.
>>
>> Thanks,
>>
>> Avi
>>
>>         [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-SIG-windows mailing list
>> R-SIG-windows at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-windows