fix(gallery): give masonry tiles their real shape back (#1130, #1131)

Two independent causes of the same symptom — an aspect-ratio layout that does
not lay anything out.

gallery-premium discarded the tile height MasonryPhotoAlbum computed from
photos.width/height and set height:auto on both card and image, so the rendered
shape came from whatever rendition was served. With thumbnail_fit seeded 'cover'
every rendition is square, so the masonry drew identical squares.

The bundled CSS templates pinned images to a fixed pixel height, which beats the
.h-full utility six of the seven layouts rely on. Elegant Dark is seeded
is_default, so that was the out-of-the-box result for any layout other than
grid/timeline.

Migrations 052/053 corrected for fresh installs; 175 repairs the rows already
seeded. The repair is whitespace-tolerant because sanitizeCSS strips newlines
from any template ever saved through the editor, matches the height property
with a lookbehind so line-height/max-height are untouched, handles grouped
selectors and skips nested rules.

Stable twin of #1135.
This commit is contained in:
Paul Nothaft
2026-08-22 21:36:51 +02:00
committed by GitHub
parent da44f1947b
commit d977e3e296
5 changed files with 388 additions and 9 deletions
@@ -77,7 +77,11 @@ const DEFAULT_CSS_TEMPLATE = `/*
.photo-card img {
width: 100%;
height: 200px;
/* 100%, not a fixed pixel height: every aspect-ratio layout (masonry,
justified, mosaic, gallery-premium) gives .photo-card a definite height
computed from photos.width/height, and this rule's specificity (0,1,1)
beats the .h-full utility (0,1,0) the layouts rely on — #1131. */
height: 100%;
object-fit: cover;
transition: transform 0.3s ease;
}
@@ -503,7 +503,9 @@ const LIQUID_GLASS_DARK = `/*
.photo-card img {
width: 100%;
height: 240px;
/* See #1131: a fixed height here beats the layouts' .h-full utility and
detaches the image from its aspect-ratio-sized card. */
height: 100%;
object-fit: cover;
transition: transform 0.4s ease, filter 0.4s ease;
filter: brightness(0.9);
@@ -639,7 +641,7 @@ const LIQUID_GLASS_DARK = `/*
}
.photo-card img {
height: 180px;
height: 100%;
}
/* Reduce animation complexity on mobile */
@@ -0,0 +1,123 @@
/**
* The bundled CSS templates pinned every gallery image to a fixed pixel
* height, which broke every aspect-ratio layout (#1131).
*
* Six of the seven layouts size a tile by putting a computed pixel height on
* `.photo-card` and letting the image fill it with `h-full`. A template rule
* of `.photo-card img { height: 200px }` has specificity (0,1,1) and beats
* `.h-full` at (0,1,0), so the image detached from its card: masonry rendered
* correctly-shaped cards with a 200px image glued to the top and empty
* background below — or, where the computed card was shorter than 200px, an
* image taller than its own container.
*
* "Elegant Dark" is seeded `is_default = true`, so this was the out-of-the-box
* result for anyone choosing any layout other than grid/timeline (where a
* fixed square happens to look deliberate).
*
* Migrations 052 and 053 are corrected for fresh installs; this repairs the
* rows already seeded. Templates are referenced by `events.css_template_id`
* and read at serve time rather than copied onto the event, so fixing the row
* fixes every gallery using it.
*
* SCOPE: every `.photo-card img` rule that carries a fixed PIXEL height, in
* every template — not just the two we seeded, and not just their pristine
* copies.
*
* That is broader than it first looks, and deliberately so. It is also not the
* scope this started with: matching the exact seeded text missed every install
* where the template had ever been saved through the editor, because
* `sanitizeCSS` strips newlines. Those are the majority, and a migration that
* silently no-ops on them while being recorded as applied is worse than none.
*
* The cost is that a fixed pixel height a user wrote themselves is rewritten
* too. That is judged acceptable because there is no layout it can be right
* for: all seven give `.photo-card` a definite height and expect the image to
* fill it, so a pixel height on the image can only detach it from its card.
* Anything that is not a fixed px height — %, vh, auto — is left alone, as is
* every declaration outside a `.photo-card img` body.
*/
/**
* Every `.photo-card img { … }` rule body, however it is spaced.
*
* Matching the exact seeded text does NOT work, and the reason is worth
* stating: `sanitizeCSS` strips all control characters (cssSanitizer.js:61),
* so the moment an admin saves a template through the editor — even only to
* rename it or toggle it — every newline is REMOVED from the stored CSS. The
* shipped `.photo-card img {\n height: 200px;` becomes
* `.photo-card img { height: 200px;`. An exact-match migration would find
* nothing on those installs, be recorded as applied, and leave the galleries
* broken with no second chance.
*
* Scoped to the rule body rather than the whole stylesheet, so the other pixel
* heights in these same templates — a 1px gradient divider, an 8px scrollbar —
* are untouched.
*/
/*
* Two details in this pattern are deliberate:
*
* * the selector part is a LIST, so `.photo-card img, .thumbnail img { … }`
* is recognised. Requiring `{` straight after `img` skipped grouped
* selectors entirely — and the migration would still be recorded as
* applied, so the template kept the bug with no second chance.
*
* * the body excludes braces, so a rule containing a NESTED block is not
* matched at all. `.photo-card img { & + .caption { height: 200px } }` is
* valid, passes the validator, and a `[^}]*` body would have captured the
* nested block and rewritten the caption's height instead. Skipping it
* means such a template keeps a fixed image height; corrupting unrelated
* declarations in a migration that cannot be undone is the worse of the
* two, and nesting does not appear in anything we ship.
*/
const PHOTO_CARD_IMG_RULE = /([^{}]*\.photo-card\s+img[^{}]*)\{([^{}]*)\}/g;
/**
* Only a fixed PIXEL height is wrong here; %, vh, auto and the rest stay.
*
* The lookbehind is load-bearing rather than defensive: without it the pattern
* matches the TAIL of `line-height`, `max-height`, `min-height` and any custom
* property ending in `-height`, and silently rewrites those instead — in a
* migration whose down() is deliberately irreversible.
*/
const FIXED_PX_HEIGHT = /(?<![\w-])height\s*:\s*\d+(?:\.\d+)?px/gi;
function relaxFixedImageHeights(css) {
return css.replace(PHOTO_CARD_IMG_RULE, (whole, selectors, body) => {
// .test() on a /g regex advances lastIndex, so it is reset on both sides
// of the check — leaving it set makes the NEXT rule start matching from an
// arbitrary offset and silently skip declarations.
FIXED_PX_HEIGHT.lastIndex = 0;
if (!FIXED_PX_HEIGHT.test(body)) return whole;
FIXED_PX_HEIGHT.lastIndex = 0;
return `${selectors}{${body.replace(FIXED_PX_HEIGHT, 'height: 100%')}}`;
});
}
exports.up = async function up(knex) {
if (!(await knex.schema.hasTable('css_templates'))) return;
const rows = await knex('css_templates').select('id', 'css_content');
let fixed = 0;
for (const row of rows) {
const original = row.css_content;
if (!original || typeof original !== 'string') continue;
const updated = relaxFixedImageHeights(original);
if (updated !== original) {
await knex('css_templates').where({ id: row.id }).update({ css_content: updated });
fixed += 1;
}
}
if (fixed > 0) {
console.log(` 175: relaxed the fixed image height in ${fixed} CSS template(s)`);
}
};
exports.down = async function down() {
// Deliberately irreversible. Putting the pixel heights back would re-break
// every aspect-ratio layout, and the rows may have been edited since — there
// is no version of "restore" here that is safer than doing nothing.
};