What check_page actually catches (and why we lean on it hard)
Reading the page instead of guessing
Most of the mistakes that make a generated page look broken are invisible in the source code. A link nested inside another link parses fine as HTML text but splits a card into two boxes the moment a real browser lays it out. A script that fails halfway through leaves a page stuck on "Loading…" forever, with nothing in the markup to say why. These are the kinds of problems that only show up once something actually renders the page, which is exactly what check_page is for.
What it actually checks
Every call renders the page in a real browser, at both a desktop width and a phone width, and reports back on several things at once:
- Console errors. If a script throws, syntax error or otherwise, it shows up here rather than staying silent.
- Failed requests. A typo in an image path or a script tag pointing at the wrong file gets flagged instead of just quietly failing to load.
- Horizontal overflow. Anything wider than the viewport, a fixed-width element, an un-wrapped table, gets caught before a visitor has to scroll sideways to read a page.
- Broken images. Missing files or bad paths are reported directly rather than showing up as an empty box no one investigates.
- Invalid nesting. An interactive element inside a link or button, the classic cause of a card that visually falls apart with no obvious CSS explanation.
- Uneven card heights. Grid rows where one card is taller than its neighbors for no good reason.
Alongside all of that, it returns an actual screenshot, so a layout problem can be looked at directly instead of inferred from a wall of CSS.
A real example
While working on this very site, a script file ended up truncated mid-function, a genuine syntax error. The homepage was stuck showing "Loading posts…" forever, and reading the source alone made it look fine at a glance since the truncation was easy to miss. Running check_page against the page surfaced the console error directly: "Uncaught Unexpected end of input". That one line pointed straight at the real problem, a cut-off script, rather than something cosmetic like lazy-loaded images still loading in.
Why this matters for anyone building without touching code
The point of a tool like this isn't to replace careful building, it's to catch the class of mistake that looks completely fine in the code and only breaks once a real visitor loads the real page on a real screen. Checking after every layout change, and especially whenever something looks "scattered" or "wrong" on a live preview, means those problems get caught immediately instead of shipping quietly to visitors.

Comments
Loading comments…