This should mitigate issues with mapped /code not owned by UID 1000.

Previously, the version file's name was supposed to be the version number, too.
This change allows version files to be named "branch_version", for example, as
the actual version number gets pulled from the file.

To use the dockerized build, replace ./build.sh with ./docker-build.sh in your commandline.
The script constructs a container and starts it. In the container several environment parameters
are fixed:
  * the base OS: Ubuntu 14.04 LTS
  * the hostname: "ffpb-build"
  * the source directory: /code

In theory, this should be enough to get the same build output no matter where the firmware gets
compiled and/or by whom.

preparation (only necessary if Dockerfile changes):
docker build -t ffpb/build .

usage (example):
docker run --env BRANCH=stable BASE=v2014.4 VERSION=0.6.1 ffpb/build

The 'V=s' appendix to calls to 'make' enable verbose output which is
useful for debugging build issues.

Previously, the locally-checked-out master got checked out which was
obviously not the intended behaviour.

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

Without testing whether a build-info.txt already exists in
./output/${BRANCH}, all build-related variables simply were appended to
such an existing file, which of course does not yield a build-info.txt
in the desired format. Hence, it becomes necessary to delete outdated
build-info.txt files prior to creating a new one.

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

builds.

As HEAD is a "floating" pointer, it cannot act as valid qualifier for
reproducible builds. One either has to specify a commit by its tag (if
existent) or the qualifying commit SHA.

By now, the value of the DATE field in the manifest header was set to be
equal to the faketime timestamp we use for the build. However, as it
turned out with the rollout of 0.6.0~rc2, the Gluon autoupdater relies
on a slightly different command and thus rejected the frankenmerged
manifest file. By "different" I mean that it is identical up to the last
character, but requires a suffix which indicates the UTC offset. E.g.
when the faketime timestamp is "2015-02-14 23:42:00", the correct value
in the DATE field looks like "2015-02-14 23:43:00+01:00", Hence, the
issue gets fixed by appending a simple "$(date +%:z)" to the TS variable.

By now, 'make dirclean' got executed for branches not equal to 'exper-
imental' no matter if one just cloned all the repos or tries to rebuild
an image. Hence, in the former case where there is nothing to clean,
make complains that 'make dirclean' fails as $GLUON_BUILDDIR does not
exist yet.

Thus, we now test whether or not the directory exists before calling
'make dirclean' and only call it, if it does not exist (i.e. one just
cloned the repos).

Keep in mind that 'make dirclean' still does not get called in case BRANCH is
set to 'experimental'.

Closes issue #20.

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

As build-info.txt contains all necessary information to produce clones
of the images that get build in a particular run, obtaining this file at
an earlier stage of the build process avoids that others will have to
wait for the result of the build process to generate, for instance, a
verification build on a different machine.

As it was before, the file gets written to "./output/${BRANCH}/", but
still only is saved in "./versions/" with its version-name if the build
process finishes successfully.

This closes issue #21.

Signed-off-by: Stefan Laudemann <thisco@webcake.de>

The variable takes a space separated list of platforms for which images
should be created. If unset, images for all target platforms in
${CODEDIR}/targets/ will be build.

If the value is not specified by the user explicitly, the value is chosen to be
equal to the number of CPUs/cores in the users machine.