Is there a problem that this is solving? The current bash -xec gives nice log output when building.

I didn't notice a big difference when I tested, but for me it feels like a hack, a way around the Dockerfile format. For example the use of cd instead of WORKDIR. Since this is an official language repo I thought it would be nice to show how it's supposed to be done, on the other hand both approaches do work and why change something that isn't broken?

Using WORKDIR would be nice to do if we had multiple RUN lines that needed it and wanted to leave it for users of the image. In this case we would need a second WORKDIR to reset it after we finish with the RUN, so we would just be adding layers.

I hadn't thought about the extra layers. So I removed the WORKDIR, and currently it removes the bash -xec '...'. This makes it easier to read IMHO, but will still bail if something goes wrong using || exit $?.

Ok, that's fair, but how about this, so the spirit of the original is kept?

RUN set -ex; \
    cd /usr/src/go/src; \
    for platform in $GOLANG_CROSSPLATFORMS; do \
        GOOS=${platform%/*} \
        GOARCH=${platform##*/} \
        ./make.bash --no-clean 2>&1; \
    done

The set -ex will both make it so that the entire thing exits nonzero if one command does, and will also print out each command as it's executed (which is really nice for giving us progress of which platform we're currently building for, since it takes a really long time to build them all).