[Pkg-phototools-devel] Bug#815712: ilmbase: FPU functions unimplemented on kfreebsd/hurd

[Steven Chamberlain]

The next test failure is a bit more tricky to understand:

| ImathTest: testBox.cpp:286: void {anonymous}::testExtendByPoint(const char*) [with T = Imath_2_2::Vec2<float>]: Assertion `b.min == min && b.max == max' failed.
| [...]
|         b = {min = {x = -7524.7915, y = <optimized out>}, max = {x = -7524.7915, y = 5073.8335}}
|         min = {x = -7524.7915, y = 5073.8335}
|         max = {x = -7524.7915, y = 5073.8335}

I tried to implement an approximatelyEqual() function for it but
there was another failure after that, in ImathTest/testBox.cpp
testExtendByPoint():

         IMATH_INTERNAL_NAMESPACE::Box<T> b;
         b.extendBy(p);
-        assert(b.min == p && b.max == p);

|         p = {x = 2970.1700842380524, y = -6529.1898965835571}
|         b = {min = {x = 2970.170166015625, y = -6529.1898965835571}, max = {x = 2970.1700842380524, y = -6529.1898965835571}}

A difference of 0.000081778 sounds quite big for double-precision.
I'm starting to think there is really a bug somewhere.

In fact, I think both code examples above are meant to simply assign
min = max = b, so there should be no error at all?

I suspect this is related to gcc-5_5.3.1-7 -> gcc-5_5.3.1-9
and affects only specific 32-bit arches.

I was testing on kfreebsd-i386;  I don't know for sure the failures
above happen also on linux-i386.

Regards,
-- 
Steven Chamberlain
steven at pyro.eu.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-phototools-devel/attachments/20160223/bc3d9caa/attachment.sig>