For unknown reasons the netlify pipeline ends up in a state where a git
worktree is not created after a fresh clone of the repository. The
previous implementation checked for the actual checked out working tree
and assumed the administrative git worktree files have to exist as well
in that case. This change explicitly verifies the existence of both path
structures and handles more cases of inconsistencies gracefully.
Signed-off-by: Tobias Schmidt <tobidt@gmail.com>
So far every prometheus/alertmanater/... release branch had to be
manually configured in the nanoc.yaml config file. With this change the
most recent release branches will be checked out automatically if a
corresponding semver tag exists.
As the Prometheus git repository includes several hundreds of megabytes
of vendored assets, the repository is cloned bare and all blobs are
filtered by default. Each version is then checked out in an individual
working tree and git's spare-chekcout feature is used to reduce the
checkout to the `docs/` folder. The git data is cached in
`tmp/repo_docs/` and will be recreated automatically if removed.
Signed-off-by: Tobias Schmidt <tobidt@gmail.com>
* Update Ruby gems.
* Remove nanoc-cachebuster due to incompatibility.
* Add README item about GitHub auth to improve build reliability.
* Fix various required config changes for 4.0
Signed-off-by: Ben Kochie <superq@gmail.com>
With this commit it becomes possible to define additional data sources
for content pages. The content will be cloned from the specified
repository. Links in these pages will be normalized to work inside the
docs project. It's assumed the documentation pages include the standard
nanoc config options and set :title and optionall :sort_rank.
