This guide aims to help developers working on this repository.
The container images created from this repository contain several runtime scripts
that are executed from the image entrypoint / default command. These scripts are
located under ${IMAGE_VERSION}/root/
.
As a suggestion, start making changes to files of a single image version, e.g., the latest one, and then run tests for that image:
make test VERSIONS=3.2 SKIP_SQUASH=1
By default, the last step in the image build process is to squash the output image into less layers, optimizing image size for distribution. However, unless specifically testing that this squash process is working, you don't need to squash images in development, and can save some precious time with consecutive rebuilds as you work in the image contents.
The SKIP_SQUASH=1
option disables image squashing, so that building and
rebuilding images locally should take less time.
Setting VERSIONS=3.2
equally saves development time. Instead of building every
image version, only the version you are working on will be built.
The tests run via make test
cover some basic usage patterns of the images.
Often times, some manual testing is required while making changes to the scripts
that are copied into the images. For instance, you can build and rebuild a
specific image (without running tests) with this command:
make VERSIONS=3.2 SKIP_SQUASH=1
When you are done with your changes, you probably want to change all other image versions without copy-paste.
We keep the multiple files for the different image versions in sync. Here is a
tip on how to use git
commands to automatically apply patches targeting one
image to all the others, without incurring in manually editing each copy of each
modified file.
Supposing that changes were made to files in the 3.2 directory, we can then apply those changes to all other images:
for version in 2.4 2.6 3.0-upg; do
git diff -- 3.2 | git apply -p2 --directory ${version} --reject
done
Depending on the changes and the surrounding context, the patch may not apply
cleanly. In that case, Git will create *.rej
files with the changes that could
not be applied.
You can show them all with:
find -name '*.rej' -exec cat {} \;
Fix the differences manually, then delete the *.rej
files.
Notes:
- Sometimes it may be useful to ignore part of the context passing the
-C<n>
flag togit apply
. - Read
git help diff
andgit help apply
for details about the flags used here and further insights on how to combine them.
Review the changes before committing to make sure the patch to all versions make sense.
Sometimes it is also useful to use a graphical diff tool such as Meld for a final verification, or working side-by-side on the parts that actually differ and that cannot the copied over automatically.