[foxis]

Fixes the CI failure introduced by #33, and resolves a conflicting-signal defect it also introduced.

1. The build break

render of "/404" failed: ... at <.Paginator>: error calling Paginator: pagination not supported for this page

hooks/head-end.html renders from the <head> in the theme's baseof.html. .Paginator hard-errors on any page kind that cannot paginate, and /404 is one. Reproduced locally with hugo 0.163.3 extended, byte-identical error.

Guarding the call by kind is worse than removing it. head-end runs before the main block, so touching .Paginator initialises pagination before the theme's archive.html calls .Paginate $pages with its own page set:

build	paginator pages (en / lt)
pre-#33	101 / 29
kind-guarded .Paginator	135 / 32

The 34 extra pages are the home paginator switching from mainSections (3 pager pages) to all RegularPages (30, stuffed with snippets), plus spurious /page/1/ duplicates on snippets/ and all eight snippet subsections in both languages.

So the clause is removed, not guarded. A comment in the partial records why, so the test does not get re-added.

Pager pages need no special handling anyway: Hugo re-renders the same Page for each pager, so .Kind and .Data.Plural hold on /tags/page/7/ and /tags/fpv/page/3/ too. The 6 URLs not covered that way (page/2, page/3, lt/page/2, fpv/page/2, lt/fpv/page/2 and its en counterpart) are left indexable deliberately: hugo 0.163 exposes no read-only pager accessor, and .Permalink does not vary between pager renders.

2. Tags only, from one shared list

#33 noindexed taxonomy and term for all three taxonomies. The sitemap override in f8d0102 keeps categories and series listed on purpose. Its own header comment:

Categories (22) and series (5) are curated and stay listed.

So 47 sitemap URLs were being advertised and noindexed simultaneously: 37 category pages, 10 series pages.

Narrowed to the tags taxonomy, driven by a single param so the templates cannot drift apart again:

thinTaxonomies = ["tags"]   # config/_default/params.toml

• layouts/sitemap.xml excludes it from the <urlset>
• hooks/head-end.html serves it noindex, follow

Both default to slice "tags" if the param goes missing, so a config edit cannot silently un-noindex or re-list the tag archive.

Scope note: this is index hygiene, not crawl reduction

noindex is an indexing directive. Googlebot must fetch the page to read it, and follow keeps it traversing. robots.txt stays Allow: / deliberately, because combining Disallow with noindex on the same URL means the noindex is never seen.

The crawl-budget lever was already f8d0102, and it works. Measured: 578 sitemap URLs covering 531 of 533 real content pages, zero tag pages. The two absent are home page/2 and page/3, which is correct.

Verification

Full local build, hugo 0.163.3 extended, same flags as CI:

• build completes, /404 renders, paginator pages back to 101 / 29, identical to pre-seo: noindex taxonomy, term, and paginated pages #33
• noindex, follow present on /tags/, /tags/<term>/ and their pager pages, both languages, 1292 pages, and nothing outside /tags/
• noindex absent on /404, /, /page/2/, /categories/, /categories/fpv/, /categories/fpv/page/2/, /series/, /fpv/, /fpv/page/2/, /lt/, /lt/categories/, /snippets/, all regular posts
• sitemap 578 URLs: 0 tags, 37 categories, 10 series, and 0 of the 578 carry a noindex tag

Two verification traps worth recording, both of which produced a false all-clear first time:

• --minify strips attribute quotes, so the output is <meta name=robots ...>. A quoted-attribute grep matches nothing and reports zero noindex tags on a working build.
• /sitemap.xml is a <sitemapindex>; the URLs live in /en/sitemap.xml and /lt/sitemap.xml. Reading the root file alone makes the sitemap look near-empty.