Description of problem:

When compiling the latest development version of darktable (from the
git repo at https://github.com/darktable-org/darktable), the Fedora
22 version of gcc 5.1.1 errors out with an apparently spurious error:

  /data/code/git/darktable/src/common/interpolation.c: In function ‘dt_interpolation_compute_sample’:
  /data/code/git/darktable/src/common/interpolation.c:789:21: error: array subscript is above array bounds [-Werror=array-bounds]
         h += kernelh[j] * ipixel[0];
                     ^
  cc1: all warnings being treated as errors
  src/CMakeFiles/lib_darktable.dir/build.make:829: recipe for target 'src/CMakeFiles/lib_darktable.dir/common/interpolation.c.o' failed

At the moment the specific line of source code producing this is:
https://github.com/darktable-org/darktable/blob/b2219b57bd2bbb56c28db390f3eec9eb2b4dad1e/src/common/interpolation.c#L789

As far as I can see, this warning is spurious and inaccurate. It did not
trigger in the Fedora 21 gcc 4.9.2, and clang produces no warnings here.

The darktable developers have pointed me at the following upstream
gcc issues as potentially relevant:

	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53198
	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=45180

Version-Release number of selected component (if applicable):

gcc-5.1.1-4.fc22.x86_64

How reproducible:

Completely.

Steps to Reproduce:

	sudo dnf builddep darktable
	mkdir /tmp/scr
	cd /tmp/scr
	git clone https://github.com/darktable-org/darktable
	cd darktable
	git checkout b2219b57bd2bbb56c28db390f3eec9eb2b4dad1e
	./build.sh -j 1

(The git checkout is optional; it is the current git tip, which I
know reproduces this.)

Actual results:

The above error is produced.

Expected results:

Darktable is compiled successfully (in /tmp/scr/darktable/build).