The loose ends: heading levels, the event page tags, a check, and the /events/ listing

[DZPM]

The loose ends the earlier pull requests of this phase left behind. Four of them, each a commit, each independent of the others.

1. 14 pages skip a heading level

The HTML validator reports it as an error, and it makes a screen reader read the wrong outline: a reader moving by heading gets no way back up from a subsection that was never under a section.

Six sources, not one:

Source	Was	Is	Pages
bordered-col, the shortcode behind the cards	h4 under the h1	h2 class="h4"	3
contact.html	two h4 under the h1	h2 class="h4"	1
options.html	h4 under an h2	h3 class="h4"	7
home.html	h4 under an h2	h3 class="h4"	1
Six event pages	18 #### under a ##	###	6
promote-your-event, the 2017 PyData page, the 2018 PyDay page	### and raw <h3> under the h1	h2	3

Every one keeps the size it had, through the h4 and h3 Bootstrap classes, which is the pattern event.html already uses with h2 class="h1". Nothing moves on the page: only the level changes, and the level is what carries the outline.

2. 41 of the 48 validator errors on PyDay BCN 2025 are a typo in a template

Not in the page, in event_detail.html: it wrote <b>Requirements:</b></b> and <b>Repository:</b></b>, with the closing tag doubled, which is 30 stray end tags across the 15 workshops of that page. It also wrapped markdownify in a p, and markdownify emits a p of its own, so every description nested one inside another: 6 more.

</br> is not a tag. A br is void and has no closing form, so a browser reads </br> as an opening one and the page gets a line break nobody asked for. 36 of them across the 2022, 2023, 2024, and 2025 pages.

And one that is not a markup problem but a broken link a reader can press: the Registration section of PyDay BCN 2025 says "General registration will open ... through Eventbrite" with href=''. An empty href reloads the page. It now carries the URL the Important dates section of the same page already has, which answers 200.

The 7 errors this does not fix are the <style> the agenda writes into the body. That one is not a typo: the agenda builds its grid from the number of tracks and the times of each page, so the rule cannot be static. Moving it needs custom properties on the container and the rest of the grid in agenda.scss, which is a rework of the schedule and not a tidy-up.

3. A check, so the fifteenth page cannot happen

Nothing stopped the heading skips coming back. bin/check-rendered now fails a build that has one.

It goes there and not in check-content because a heading comes from three places at once: the markdown of a page, a shortcode, and the template around both. Only the built page knows the order they ended up in, which is why the 14 were six different sources.

It reads the main element with the person modals taken out, because a modal holds a heading tree of its own inside a dialog that is closed until a reader opens it. The frozen copies under static/archives/ are skipped, the same exemption the http check already makes.

Checked both ways: nothing is reported on the build as it is, and putting one heading back to an h4 in contact.html makes it report contact/index.html h1 -> h4 and fail.

4. /events/ had no listing

It answered 200 with the word "Events" and nothing under it, which is the open half of an item of #190. #202 took /people/ away because nothing linked to it and it held nothing; /events/ is a real section with real children and the fix is the opposite.

One line. layouts/_default/list.html already lists a section's pages, and content/events/_index.md asked for layout: "single" so the fallback never ran. The page now lists Monthly Events, Other events, PyDataBCN, and PyDay BCN, each linked and with its description.

---

This closes the last two open items of #190, and the first three commits are each measured against the build rather than asserted.

[ifosch]

LGTM! Great addition to review header skip levels... Looking forward to see how the events listing looks like. Thanks!