Bug#810859: postgis FTBFS on hppa and mips architectures (with patch)

[Helge Deller]

Hi Bas,

On 13.01.2016 16:15, Bas Couwenberg wrote:
> Your patch has triggered some discussion in the upstream issue
> tracker, can you join that discussion to answer the questions from
> the upstream developers? Specifically this question: "is the IEEE NaN
> number (as you can see in the test) a valid [finite] double on
> mips/hppa?"
> 
> https://trac.osgeo.org/postgis/ticket/3426

Since I have to run in a few minutes (and since I don't have an account there),
could you please paste my answer in the ticket?
Answer:

> This change probably also fails to address a number of other big-endian platforms, like sparc64 and BE arm variants. 

I don't think so. When you look at the debian build results on the various architectures, it seems that
only hppa and mips seem to be affected by this NaN problem (sparc64 is OK if you click on "old" in sparc64 line):
https://buildd.debian.org/status/package.php?p=postgis&suite=sid

> Sebastiaan, is the IEEE NaN number (as you can see in the test) a valid [finite] double on mips/hppa ? 

I don't know.

Regarding the C-testcase I get on parisc/hppa:
root at phantom:~# ./testnan 
nan is nan ? 1
nan is nan ? 1

If someone needs/want a ssh-login on a parisc machine, just let me know.

Thanks,
Helge