Script: HTML `_static/` gets 225 macOS AppleDouble files (`._*`) from the Runestone Services archive

[rbeezer]

Summary

Every HTML build that places Runestone Services in its own output (the default; not portable
HTML, nor resources from the CDN) ends up with 225 macOS AppleDouble
files in _static/, one beside each real file:

_static/._139.6404c46752b3e634.js
_static/._139.6404c46752b3e634.js.gz
_static/._139.6404c46752b3e634.js.map
...

Each is 163 bytes of Finder metadata (file reports "AppleDouble encoded Macintosh
file"), used by nothing. They are harmless in a browser, but they double the file count
of _static/ (450 files where 225 are wanted), and they travel with every site that is
uploaded or zipped from the output.

Cause

They are in the archive itself. dist-8.3.2.tgz from https://runestone.academy/cdn/runestone/
has 450 entries, 225 of them ._*, all owned by bmiller/staff; dist-8.2.13.tgz is the
same. This is what macOS tar does by default: it writes each file's extended
attributes as a ._ companion. The archive also carries LIBARCHIVE.xattr.com.apple.provenance
headers, which GNU tar reports as "Ignoring unknown extended header keyword" on every
entry; Python's tarfile passes over those silently.

_place_runestone_services (pretext/lib/pretext.py, lines 3958–3959 at 20d10545)
unpacks everything:

            services_file = tarfile.open(services_build_path)
            services_file.extractall(build_dir)

Suggested fix

Either end would do, and both would be best.

In the script, skip the AppleDouble members when extracting:

            services_file = tarfile.open(services_build_path)
            members = [m for m in services_file.getmembers()
                       if not os.path.basename(m.name).startswith("._")]
            services_file.extractall(build_dir, members=members)

Tried on dist-8.3.2.tgz: 225 files come out, none of them ._*, and they are exactly
the files a build places now, less the AppleDouble ones.

In the Runestone packaging, make the archive without the macOS metadata, for example
COPYFILE_DISABLE=1 tar --no-xattrs -czf dist-<version>.tgz ..., which also removes the
provenance headers. That fix belongs to the Runestone side.

Reproduce

Any HTML build with the default static files, at 20d10545:

pretext -c doc -f html -p publication.ptx -d out main.ptx
ls -A out/_static | grep -c '^\._'        # 225

A related observation

extractall() without a filter argument raises a DeprecationWarning on Python 3.12
and later ("Python 3.14 will, by default, filter extracted tar archives..."); it is not
seen because the warning comes from a library module. Passing filter="data" where
tarfile has it (3.12, and the security releases 3.10.12 and 3.11.4) would settle that
at the same spot.