/* Written by _scripts/o45_legibility_pass.py.

   THE TAP-TARGET OVERLAY WAS TRIED THREE WAYS AND REMOVED ALL THREE TIMES. The client asked for
   finger-sized targets ("24 is probably a bit small too"), and the obvious mechanism - an `::after`
   band 44px tall, which grows the hit area without moving anything - was measured on al-vita.com
   and makes things WORSE wherever links stack tightly. A submenu at a 26px pitch had its anchors
   fall from owning 26 of 26 hit rows to 14; excluding nested lists moved the damage to the footer,
   where the email link fell from 26 rows to 4. Neighbouring bands overlap, and the browser gives
   the shared rows to whichever painted last. The blocking count went 6 -> 9 -> 9 while the site
   itself never changed. A remedy that removes rows from real controls is not a remedy.

   Growing them honestly means padding, and padding on one shared rule across eighteen different
   themes moves headers about. That is per-site work where somebody can look at the result, and it
   is written down as such rather than done blind. What stays here cannot misfire. */

/* A link inside prose that shares its paragraph's colour, weight and lack of underline is not a
   link a reader can see. Underline restores the affordance without touching colour, size or
   layout - the oldest convention on the web, and the one the readability complaints point at.

   PARAGRAPHS ONLY. The first version also matched `li a` and `td a`, and a navigation menu is a
   list: al-vita.com's entire top nav came out underlined - SOBRE NOS, SERVICOS, APRESENTACAO DO
   ESPACO - which is exactly the kind of thing the client means by "it catches the eye". Nothing in
   CSS can tell a prose list from a menu, so the rule stays where the distinction is safe. */
p > a:not([class]), p a:not([class]):not(nav a):not(header a):not(footer a) {
  text-decoration: underline;
  text-underline-offset: 2px;
}