Skip to content

networkupstools/nut-website

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Network UPS Tools website

This repository contains the scripts needed to generate the NUT website.

Since it was originally part of the NUT source tree, it shares some history, and you can save some bandwidth if you already have a copy of the NUT repository downloaded:

:; git clone --reference /path/to/nut \
    https://github.com/networkupstools/nut-website.git

Once you have cloned the nut-website repository, you can initialize the submodules, and pull your copy of NUT into the website tree as well:

:; git submodule init
Submodule 'ddl' (https://github.com/networkupstools/nut-ddl.git) registered for path 'ddl'
Submodule 'nut' (https://github.com/networkupstools/nut.git) registered for path 'nut'
:; git submodule update --reference /path/to/nut nut
...

Required Packages

You will need a copy of AsciiDoc toolkit, a2x (part of AsciiDoc), and its dependencies for MAN, HTML and PDF document formats generation (including dblatex, xmllint, and xsltproc). Current version requirements will be listed in the output of the ./configure script. To build the Hardware Compatibility List (HCL) page, you will need either the simplejson or json Python module (the json module that comes with Python 2.7 will work) and the lxml module. You will also need autoconf and automake, and possibly libtool since the NUT module uses it.

Note

With recent Ubuntu/Debian releases, Python2 is deprecated so much that pip2 tool seems to be no longer packaged. If you have both python2 and python3 installed, you may have install modules by pip3 or APT and to either declare the preference via environment variables for NUT and NUT-Website configure and other scripts, or change system-wide default symlink to python once:

:; sudo apt-get install python3-pip
:; python3 -m pip install lxml simplejson pycparser pathlib

### Note: Newer Python releases can suggest to use APT packages for modules
### too (note also that "pathlib" may be not available this way, and may be
### in fact part of the base Python distribution); in this case:
:; sudo apt-get install python3-{lxml,simplejson,pycparser}

and then either system-wide:

:; sudo apt-get install python-is-python3

or constrain the preference to nut-website builds:

:; PYTHON='/usr/bin/env python3' ./ci_build.sh

The source-highlight package is optional, but if available, will be used by AsciiDoc for syntax highlighting of examples.

The optional htmlproofer tool from https://github.com/gjtorikian/html-proofer project can be used to sanity-check links and similar aspects of the markup in generated HTML pages. On Debian/Ubuntu systems you can install it as a package:

:; sudo apt-get install ruby-html-proofer
Note

If your htmlproofer runs complain like this:

htmlproofer 3.19.2 | Error:  "\xC3" on US-ASCII
  /usr/lib/.../nokogiri/html5.rb:389:in `encode': "\\xC3" on US-ASCII (Encoding::InvalidByteSequenceError)

Try exporting LANG and LC_ALL environment variables to use UTF-8 capable locales (already handled in the Makefile.am targets by HTMLPROOFER_ENV); this may further require installing some or all locale packages, e.g.:

:; sudo apt-get install locales-all

GNU make and GNU coreutils are recommended, but if you see any remaining non-portable constructs in the Makefiles, please let us know.

Building

Editing the Makefile.am source

Warning
Due to lines which # HIDE FROM AUTOMAKE # some GNU syntax which conflicts with automake syntax, you MUST use autogen.sh to re-generate the practical Makefile during development cycles. If you rely on automatic typical regeneration of Makefile.amMakefile.inMakefile, the resulting file can have crucial parts commented away.

Alternately, make unhide-from-automake after edits, e.g. to fiddle with spell checker Makefile recipes for historic releases:

:; make unhide-from-automake ; make spellcheck NUT_SPELL_DICT=nut-website.dict

Quick builds for CI and developer iterations

Main rituals for re-builds should now be handled well by a single script:

:; ./ci_build.sh

For maintainers (or CI agents) who may push the web-site codebase, further envvars may be useful to commit changes locally, and to push index upwards:

# To publish automatically use:
:;   export CI_AUTOCOMMIT=true
:;   export CI_AUTOPUSH=true

# Optionally (on CI) to avoid rebuilds in cases when Git sources did not
# change after a pull of all nut-website and submodule HEADs (returns exit
# code "42" then, to be handled by the caller):
:;   export CI_AVOID_RESPIN=true

# Optionally - for the rare historic-release sub-sites (by NUT tag), e.g.:
:;   export NUT_HISTORIC_RELEASE=v2.7.4

Also, to make sure that syntax of nut-ddl data dump files is not ambiguous, you can tell their parser to abort in case of doubts:

:;   export NUT_DDL_PEDANTIC_DECLARATIONS=True

:; ... make ...
Traceback (most recent call last):
  File "/home/jim/nut-website/./tools/nut-ddl.py", line 1532, in <module>
    commentsMap[pattern](comment)
  File "/home/jim/nut-website/./tools/nut-ddl.py", line 400, in nds_dev_comment_block
    raise RuntimeWarning (msg)
RuntimeWarning: Invalid device block comment: does not end with DEVICE:EOC (blank non-comment lines mid-block?)
make: *** [Makefile:1012: ddl/PowerWalker/PowerWalker__VI_2200_SH__usbhid-ups__2.7.4__01.dev.txt] Error 1

Manual building in detail

:; ./autogen.sh && ./configure && { make -k ; make ; }
Note
There are currently some issues with parallel builds (e.g. protocols sub-directory should be built before OUTDIR but target dependencies do not say so). Please run make sequentially for the time being.

The root of the website will be in the output/ directory, if all goes well.

Building release-version sites

Every once in a while NUT has releases :) and those end up packaged in many operating system distributions or equivalent bundles of binary code, and are used by the majority of NUT users.

While the main nut-website now aims to follow the active development based on current "master" or "main" branches of the components involved, we also publish reference sub-sites with static data (man pages, configuration, compatibility information, etc.) for those people to see about setting up their practical systems properly. Such pages are marked with a note on top that they represent an old and immutable codebase which may differ from modern project state.

This also impacts some but not all pages of the nut-website as well — e.g. the stable-hcl.txt file which is a wrapper to include nut/data/driver.list info, but not all of the main website files currently (just because it is too cumbersome to partially check out arbitrary old codebase to build with current updated recipes). Likewise, general markup and footers, etc. remain from the current nut-website codebase at the time of (re-)generation.

In particular, the NUT DDL pertains to all NUT releases (reports the tested release in the filenames) so is not published separately per historic release.

Normally this should be done once per release, with a call like this:

:; export NUT_HISTORIC_RELEASE=v2.7.4
:; ./autogen.sh && \
    ./configure --with-NUT_HISTORIC_RELEASE=${NUT_HISTORIC_RELEASE} && \
    { make -k dist-sig-files || make dist-files; } && \
    { make -k ; make; }

Which would populate e.g. output/historic/v2.7.4 subdirectory that would be copied and committed into same-named path under networkupstools.github.io as detailed below. Generation of the main nut-website would also populate an index file of the "historic" subdirectory, based on historic/index.txt contents, to refer to such officially published snapshots. This index is currently maintained manually, to ensure human decision about publishing (or hiding) an historic release (especially a release candidate) vs. experimenting with that.

Note
make dist-files should update the historic release site source tarballs and related ChangeLog, news and checksum files IFF the release data was not yet there. You probably need to commit that back to "source" github repository.
Note
For hardcore maintainers, there should be a PGP/GPG key to also sign the release tarball, calling make dist-sig-files (would fail without a key).

Sanity-checking the generated HTML files

If the htmlproofer tool is installed (see above) and detected by the configure script (automatically for presence in PATH), you can explicitly call make check-htmlproofer to validate the files present in OUTDIR_BASE and/or OUTDIR (if stored separately from the base for custom or historic website builds).

You can also validate the published website repository (into which you would upload the generated OUTDIR contents) as e.g. prepared by the ci_build.sh using make check-htmlproofer-OUTDIR_PUBLISHED (optionally customize the OUTDIR_PUBLISHED environment or make variable to point to the checkout location for that repository, if not using the scripted default).

Note that this check can take about 10 minutes (especially if not disabling the referenced external site availability), so it is not done by default. You can pass custom HTMLPROOFER_OPTIONS to the make operation, if desired; consider pasting the HTMLPROOFER_OPTIONS_URLSWAP in that case.

Publishing

Note
These are internal notes for the maintainers.

The build result is published to the NUT github.io master site repository as well as news maintenance on NUT github.io latest-release site repository

Hence, the rolling master site publication is as easy as:

:; git clone https://github.com/networkupstools/networkupstools-master.github.io
:; rsync -avPHK ./output/* /path/to/networkupstools-master.github.io/

Release site publication is much less frequent. It follows the master site guideline when making a release, but only requires updating the index.html file when news.txt is updated, as noted below.

Note
Be careful to use git mergetool -y to merge the updates from a newly generated index.html into the release site, to keep intact the title (marked with comments) which specifies the type of site.
:; git clone https://github.com/networkupstools/networkupstools.github.io
:; cp -R ./output/index.html /path/to/networkupstools.github.io/
:; (cd /path/to/networkupstools.github.io/ && git difftool -y)
Note
Maybe also update the ddl and stable-hcl.html on master site, as its updates often reflect newly confirmed support of devices by existing NUT releases.

Updates

If you only have a small patch (fixing a typo or wording), don’t feel obliged to install all of the dependencies listed above just to test it. Feel free to create a pull request on this repository, or (less preferable as slower to process) send the patch as an attachment to the nut-upsdev list.

Maintainer note: Publishing became part of NUT CI farm automation in 2022, so whenever master branch sources of relevant repositories are changed, the website should not lag behind too long. Needed behavior is defined in this repository in Jenkinsfile-infra file, with job history visible at https://ci.networkupstools.org/view/InfraTasks/job/nut-website/

  • As of this writing, changes of nut-website repository should get picked up quickly thanks to "web hooks" sent by GitHub to NUT CI farm servers, and changes in NUT, NUT-DDL and other repositories involved would be evaluated every 3 hours.

  • (Re-)builds of historic sub-sites for release candidates etc. are handled manually by maintainers, to publish source tarballs as well (in nut-source repository, in the web-site, in GitHub releases page), and generally happen once per such release with a spell like this:

    :; CI_AUTOCOMMIT=true CI_AUTOPUSH=true NUT_HISTORIC_RELEASE=v2.8.0-rc3 ./ci_build.sh
  • The nut-website specific spell-checking is handled with a dynamic mix of original nut/docs/nut.dict and custom nut-website.dict.addon with key words specific to files in the website (including HTML and asciidoc markup).