html,
body {
    margin: 0;
    min-height: 100%;
}

.muf-login-body {
    min-height: 100vh;
    min-height: 100dvh;
    overflow-x: hidden;
    color: var(--muf-ink);
    background: #101a35;
    font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, Arial, sans-serif;
}

.muf-login-shell {
    position: relative;
    isolation: isolate;
    display: grid;
    min-height: 100vh;
    min-height: 100dvh;
}

.muf-login-backdrop,
.muf-login-overlay {
    position: fixed;
    inset: 0;
}

.muf-login-backdrop {
    z-index: -2;
    background-color: #24375e;
    background-position: center;
    background-repeat: no-repeat;
    background-size: cover;
    transform: scale(1.015);
}

.muf-theme-day .muf-login-backdrop {
    background-image: url('/image/muf-login-hero-day.webp');
    background-position: 45% center;
}

/* The inline style on .muf-login-backdrop, written from LoginHeroService, is
 * what actually loads; these two rules are the fallback if that attribute is
 * ever lost. They are kept pointing at the SAME files so the fallback cannot
 * quietly serve a different, smaller photograph than the service chose - which
 * is what it did until 15 Oct 2026, when this one still named the 474px file. */
.muf-theme-night .muf-login-backdrop {
    background-image: url('/image/sioux-city/south-sioux-bridge-2844.webp');
    background-position: 42% center;
}

/* -------------------------------------------------------------------------
 * EVERY HERO ON THIS PAGE IS SMALLER THAN THE SCREEN IT IS STRETCHED ACROSS.
 *
 * STATE OF THE THREE FILES, measured 15 Oct 2026 from the files themselves:
 *
 *   sioux-city/south-sioux-bridge-2844.webp  2844 x 1890  EVERY night
 *   sioux-city/lower-fourth-1920.webp        1920 x 1278  alternate days
 *   muf-login-hero-day.webp                  1536 x 1024  alternate days
 *
 * On a 3840-wide display the browser now stretches them 1.35x, 2.00x and 2.50x.
 * Before 15 Oct the night one was 474 px wide and the browser stretched it
 * 8.1x, in one bicubic step, every night.
 *
 * BOTH REBUILDS ARE REPRODUCIBLE: tools/build-login-hero.sh. Running it
 * produces the committed bytes, so the numbers in these comments can be
 * checked rather than believed.
 *
 * THE FOUNTAIN, 15 Oct. He asked: "fix image and make sure it is has great
 * resolution on 40 in monitor". 960 -> 1920, two lanczos passes with light
 * unsharp. A 3840 version was built and rejected: at 1:1 the brick mortar and
 * the carved panel go mushy at 4x while 2x holds them.
 *
 * THE BRIDGE, 15 Oct. He asked: "muf-login-hero-night.webp 474 x 315 from a
 * 474 x 315 original <- shown EVERY night use canva and fix the image so it
 * can be used". 474 -> 2844, six times, which the fountain could not take and
 * this can: it is a long exposure at dusk, so sky, river and light trails are
 * smooth gradients with almost no fine detail to lose. The step that mattered
 * was not the enlarging but DEBLOCKING first - the source is a 21 KB JPEG and
 * at 6x its 8x8 blocks become 48px squares across the sky. Full recipe and the
 * reasoning in tools/build-login-hero.sh and LoginHeroService.NightPhotos.
 *
 * CANVA WAS NOT USED. It is not reachable from this session - no connector, no
 * skill - and its upscaler is generative, which is not allowed under a named
 * photographer's credit on this page.
 *
 * STILL TRUE OF ALL THREE: an enlargement is not photography. 2844 px from a
 * 474 px original invents nothing, but it adds nothing either. A bigger
 * original from the photographer is the only real fix.
 * NOBODY HAS ASKED HIM FOR ONE.
 *
 * This comment used to say he had ruled it out, quoting him as saying "not
 * happening fix it". CORRECTED 14 Oct 2026: that quote has no primary source
 * anywhere. A search of every issue, every pull request, docs/Governance/SETTLED.md,
 * docs/Chat-History/INDEX.md and both saved chat exports finds it only in agent-
 * written comments in this file and in LoginHeroService - one session's paraphrase
 * of a four-word reply, written into source and then cited by later sessions as a
 * settled owner decision. That is the exact failure CLAUDE.md names: a guess
 * hardened into a finding inside an artifact that outlives the conversation.
 *
 * What IS measured: no full-resolution original has ever been requested,
 * promised, delivered or refused.
 *
 * So instead of pretending an upscaled photo is sharp, the page softens
 * it ON PURPOSE - and only where it is actually being stretched, which is why
 * each tier is paired with a media query rather than applied flat. Blur plus a
 * little extra scale reads as depth of field; a 4x nearest-neighbour stretch
 * reads as a mistake. The card sits in its own element, so nothing here touches
 * the text.
 *
 * BE CLEAR ABOUT WHAT THIS IS: it hides a shortfall, it does not remove one. The
 * numbers that drive it live in LoginHeroService.LoginHeroState. The day a
 * full-size file arrives, raise the width there and the softening stops by
 * itself - there is nothing here to remember to undo.
 * ------------------------------------------------------------------------- */

/* A 474 x 315 PHOTO IS NOT SHOWN ON A BIG SCREEN AT ALL.
 *
 * Owner, 14 Oct 2026, after the first attempt softened it: "The images sucks ass
 * fix it". He is right, and blur was the wrong answer - it hides a shortfall
 * rather than removing one, and a blurred thumbnail still reads as a mistake
 * next to crisp text.
 *
 * Until a bigger copy of this photograph exists - and nobody has asked him for
 * one, see the correction above - the honest fix
 * is to stop asking the file to do something it cannot: above the width where it
 * stays sharp, the photo is dropped and the brand colour carries the page. A
 * clean deep-navy field behind a white card is a deliberate design; a 4x
 * upscaled photograph is not. Nothing is faked and nothing is mushy.
 *
 * Where it IS sharp - a phone, where fitting the width puts it at 0.76x - the
 * photo stays, because there it is a real photograph of the bridge.
 *
 * The day heroes are 1536 and 960 wide and are NOT affected: they are still
 * photographs at these sizes.
 *
 * The moment a full-size bridge photo arrives, raise IntrinsicWidth in
 * LoginHeroService and this rule stops applying on its own. */
/* THE PHOTOGRAPH IS SHOWN, AND IT IS NOT BLURRED. Rewritten 14 Oct 2026 after
 * RENDERING the page at 1920 and 2560 and looking at it - which, in three
 * previous attempts at this complaint, nobody had done. Tests/SignInScreenShots
 * only ever shot 390 and 1440, so the widths being complained about had never
 * been seen.
 *
 * WHAT THE RENDER SHOWED, at 2560 x 1440:
 *   night - no photograph at all. Flat navy, the card jammed against the right
 *           edge, and the footer still reading "Photos courtesy of Chadd
 *           Goosmann" over an empty field. That is the rule below as it was.
 *   day   - his Lower Fourth photograph, deliberately blurred 3px and cropped
 *           so tightly that the street either side of the fountain is gone.
 *
 * Owner, 14 Oct 2026: "and now I need those images fixed so they can look good
 * on large monitor as well as the other images". Before that, "The images sucks
 * ass fix it" - said about the blur, and he was right about it twice.
 *
 * SO BOTH OF THOSE ARE GONE:
 *
 * No blur, at any width. A browser upscaling a 960px photograph to 2560 is
 * already soft; adding 3px of blur on top of that is not depth of field, it is
 * two mistakes. Softness from an honest upscale reads as a photograph taken on
 * a smaller camera. Blur reads as broken.
 *
 * No dropping the photograph. A deep-navy field behind a white card was
 * defended in this file as "a deliberate design", and perhaps it would be, if
 * anyone had chosen it. Nobody did: it is what is left when the image is
 * removed, and the page goes on crediting a photographer for it. Showing his
 * photograph softly is honest; showing none of it under his name is not.
 *
 * WHAT WAS LEFT, AND IT WAS NOT FIXED BY CSS: his bridge photograph was
 * 474 x 315, a 5.4x upscale at 2560, and no stylesheet could make it sharp.
 * ADDRESSED 15 Oct 2026 - the file is now 2844 wide, built from that same
 * original by tools/build-login-hero.sh, so the browser stretches it 1.35x at
 * 3840 instead of 8.1x. The scrim below is still wanted, for the white card to
 * sit against; it is no longer carrying a resolution shortfall.
 *
 * An enlargement is not photography. A full-resolution copy from the
 * photographer is still the only real fix and - measured across every issue,
 * pull request, SETTLED.md and both chat exports - has never once been asked
 * for. Raise IntrinsicWidth in LoginHeroService the day it arrives.
 *
 * Instead of blur, the scrim below carries the large-screen treatment: a
 * stronger, warmer gradient over the photograph so softness reads as evening
 * light rather than as a failure. Rendered and compared, not reasoned. */
@media (min-width: 950px) {
    /* Keep the composition. `cover` on a 1.50 photograph in a 1.78 viewport
     * scales 2.67x and crops the height, which is what threw away the street on
     * either side of the fountain. Anchoring higher keeps the sky and the
     * buildings, which are what make it read as Sioux City. */
    .muf-theme-day .muf-login-backdrop,
    .muf-theme-night .muf-login-backdrop {
        background-position: center 42%;
    }

    .muf-hero-tiny .muf-login-backdrop,
    .muf-hero-short .muf-login-backdrop,
    .muf-hero-full .muf-login-backdrop {
        filter: none;
        transform: scale(1.01);
    }

    /* The scrim does the work the blur used to. Heavier on the card side so the
     * white panel has something to sit against, light on the far side so the
     * photograph is actually visible rather than washed flat. */
    .muf-hero-tiny .muf-login-overlay,
    .muf-hero-short .muf-login-overlay,
    .muf-hero-full .muf-login-overlay {
        background:
            linear-gradient(100deg,
                rgba(10, 18, 36, 0.18) 0%,
                rgba(10, 18, 36, 0.34) 48%,
                rgba(10, 18, 36, 0.62) 100%);
    }

    /* The smallest file gets the most help, because it is being stretched the
     * furthest - a touch more scrim, still no blur. */
    .muf-hero-tiny .muf-login-overlay {
        background:
            linear-gradient(100deg,
                rgba(10, 18, 36, 0.30) 0%,
                rgba(10, 18, 36, 0.46) 48%,
                rgba(10, 18, 36, 0.70) 100%);
    }
}

/* Reduced motion means LESS MOVEMENT, not less blur, and the first version of
 * this block got that backwards: it removed the blur and kept the scale, which
 * handed the raw 4.1x stretch - the worst rendering on the site - to the people
 * least able to tolerate it, while keeping the one property that is actually
 * motion-adjacent. Corrected 13 Oct 2026 after an independent render. The scale
 * goes; the blur stays. */
@media (prefers-reduced-motion: reduce) {
    .muf-login-backdrop { transform: none; }
}

.muf-login-overlay {
    z-index: -1;
    background:
        linear-gradient(90deg, rgba(8, 19, 43, 0.08) 0%, rgba(8, 19, 43, 0.17) 46%, rgba(8, 19, 43, 0.58) 100%),
        linear-gradient(180deg, rgba(8, 19, 43, 0.05) 45%, rgba(8, 19, 43, 0.32) 100%);
}

.muf-login-stage {
    display: flex;
    align-items: center;
    justify-content: flex-end;
    width: 100%;
    min-height: 100%;
    padding: clamp(1.25rem, 4vw, 4rem);
}

.muf-login-card {
    width: min(100%, 29rem);
    padding: clamp(1.75rem, 4vw, 3rem);
    border: 1px solid rgba(255, 255, 255, 0.72);
    border-radius: 1.5rem;
    background: rgba(255, 255, 255, 0.97);
    box-shadow: var(--muf-card-shadow);
}

/* The CONDITION has to accept the prefixed form too, or this whole block is
 * skipped on iOS 17 and earlier and the prefixed line inside it never runs -
 * which is what a first pass at this got wrong. Apple's Safari 18 release notes
 * say plainly: "Added support for the unprefixed backdrop-filter. (123523441)",
 * so every iPhone on iOS 15, 16 or 17 answers false to the unprefixed test and
 * falls back to the opaque card. Owner, 13 Oct 2026: "Safari has to work we have
 * Iphone users and ipad". */
@supports ((backdrop-filter: blur(18px)) or (-webkit-backdrop-filter: blur(18px))) {
    .muf-login-card {
        background: rgba(255, 255, 255, 0.94);
        -webkit-backdrop-filter: blur(18px);  /* iOS 17 and earlier: Safari 18 was the first to take it unprefixed */
        backdrop-filter: blur(18px);
    }
}

.muf-login-card-header {
    display: flex;
    align-items: center;
    justify-content: space-between;
    gap: 1rem;
    padding-bottom: 1.5rem;
    border-bottom: 1px solid #e2e7f0;
}

.muf-login-logo {
    display: block;
    width: 9rem;
    height: auto;
}

.muf-context-pill {
    display: inline-flex;
    align-items: center;
    min-height: 2rem;
    padding: 0.42rem 0.72rem;
    border: 1px solid #d9e1ee;
    border-radius: 999px;
    color: var(--muf-navy);
    background: #f4f7fb;
    font-size: 0.72rem;
    font-weight: 700;
    letter-spacing: 0.03em;
    line-height: 1.1;
    text-align: center;
}

.muf-login-copy {
    padding-top: 1.65rem;
}

.muf-eyebrow {
    margin: 0 0 0.65rem;
    color: var(--muf-blue);
    font-size: 0.76rem;
    font-weight: 800;
    letter-spacing: 0.14em;
    text-transform: uppercase;
}

.muf-login-title {
    margin: 0;
    color: var(--muf-ink);
    font-size: clamp(1.8rem, 4vw, 2.35rem);
    font-weight: 750;
    letter-spacing: -0.035em;
    line-height: 1.08;
}

.muf-login-intro {
    margin: 1rem 0 1.6rem;
    color: var(--muf-muted);
    font-size: 1rem;
    line-height: 1.55;
}

.muf-alert {
    margin: 0 0 1.25rem;
    padding: 0.85rem 1rem;
    border: 1px solid #eab7b2;
    border-radius: 0.75rem;
    color: #7f2018;
    background: #fff1f0;
    font-size: 0.92rem;
}

.muf-microsoft-button {
    display: flex;
    align-items: center;
    justify-content: center;
    gap: 0.85rem;
    width: 100%;
    min-height: 3.35rem;
    padding: 0.75rem 1.1rem;
    border: 2px solid var(--muf-blue);
    border-radius: 0.8rem;
    color: #fff;
    background: var(--muf-blue);
    box-shadow: 0 10px 24px rgba(0, 103, 184, 0.24);
    font-size: 1rem;
    font-weight: 700;
    line-height: 1.2;
    text-decoration: none;
    transition: background-color 150ms ease, border-color 150ms ease, transform 150ms ease;
}

.muf-microsoft-button:hover {
    border-color: var(--muf-blue-hover);
    color: #fff;
    background: var(--muf-blue-hover);
    text-decoration: none;
    transform: translateY(-1px);
}

.muf-microsoft-button:focus-visible {
    outline: 3px solid var(--muf-focus);
    outline-offset: 3px;
}

.muf-ms-mark {
    display: grid;
    flex: 0 0 auto;
    grid-template-columns: repeat(2, 0.48rem);
    grid-template-rows: repeat(2, 0.48rem);
    gap: 0.12rem;
    width: 1.08rem;
    height: 1.08rem;
}

.muf-ms-mark span:nth-child(1) { background: #f35325; }
.muf-ms-mark span:nth-child(2) { background: #81bc06; }
.muf-ms-mark span:nth-child(3) { background: #05a6f0; }
.muf-ms-mark span:nth-child(4) { background: #ffba08; }

.muf-login-help {
    margin: 1.15rem 0 0;
    color: var(--muf-muted);
    font-size: 0.84rem;
    line-height: 1.45;
    text-align: center;
}

.muf-login-card-footer {
    margin-top: 1.6rem;
    padding-top: 1rem;
    border-top: 1px solid #e2e7f0;
    color: #6b7589;
    font-size: 0.76rem;
    line-height: 1.4;
    text-align: center;
}

@media (max-width: 949px) {
    .muf-theme-day .muf-login-backdrop,
    .muf-theme-night .muf-login-backdrop {
        background-position: 38% center;
    }

    /* THE PHONE IS THE ONE PLACE A SMALL PHOTO CAN BE FIXED PROPERLY RATHER THAN
     * DISGUISED, and the first version of this work missed it entirely: the blur
     * tiers above start at 950px, so a phone still got the worst of it.
     *
     * `cover` on a 474 x 315 file at 360 wide upscales it 2.58x AND crops away
     * about 70% of the frame - you see a zoomed middle slice, which is why the
     * owner's narrow screenshot reads as having no photo at all. Fitting the
     * WIDTH instead downscales to 0.76x - genuinely sharp, not softened - and
     * shows the whole bridge as a band, with the #24375e already set on this
     * element filling beneath it and the card sitting over the top.
     *
     * Measured and rendered 13 Oct 2026 at 360 x 800. This one is a real gain:
     * it stops asking the file for more pixels than it has. Must stay AFTER the
     * rule above - equal specificity, so source order decides.
     *
     * RAISED FROM 700px TO 949px ON 14 OCT 2026, because 700 left a dead band.
     * The blur tiers below start at 950, so between 701 and 949 nothing applied
     * at all: `cover` stretched the 474px photo 3-4x AND cropped it, with no
     * blur to disguise it - the exact rendering this rule exists to prevent.
     * That band is every iPhone held SIDEWAYS (844, 852, 932 - widths measured
     * in SETTLED.md:2033) and every iPad held UPRIGHT except the 12.9 inch.
     *
     * Owner, 13 Oct 2026: "Safari has to work we have Iphone users and ipad" -
     * so that band was the worst possible place to leave a gap. It is not a
     * WebKit fault: a media-query boundary behaves the same in every engine,
     * which is why it was provable without ever running Safari.
     *
     * Found because Tests/SignInScreenShotsTests only shoots 390 and 1440 and
     * never looked between them. */
    /* THIS RULE IS ABOUT THE PHONE, NOT ABOUT THE FILE - CORRECTED 15 Oct 2026.
     *
     * It used to be keyed on .muf-hero-tiny and .muf-hero-short, which are
     * QUALITY classes: they say how soft a file is, not how a frame should be
     * composed on a narrow screen. Fitting the WIDTH is wanted on a phone for
     * BOTH reasons and for every hero - a 1.50 photograph in a 0.45 viewport
     * is cropped to a ~30% middle slice by `cover` whatever its resolution,
     * which is the rendering the owner described as having no photo in it.
     *
     * Tying the two together quietly broke the fountain in 962bda1 yesterday:
     * taking it from 960 to 1920 moved it from muf-hero-short to muf-hero-full,
     * so this rule stopped matching it and its phone rendering changed from the
     * whole fountain to a cropped slice. Nothing in that commit mentioned the
     * phone, and nothing failed. Enlarging the bridge tonight would have done
     * the same thing to the night hero - 474 -> 2844 crosses the same boundary -
     * which is how it was found.
     *
     * Keyed on the theme rather than the quality class now, so a file getting
     * BIGGER can never again change how a phone frames it. Both themes are
     * named rather than using the bare .muf-login-backdrop, because that would
     * drop this to one class and lose to the .muf-theme-night rule above, which
     * sets background-position: 38% center. Equal specificity, later in the
     * file - the same arrangement the quality-class version relied on. */
    .muf-theme-day .muf-login-backdrop,
    .muf-theme-night .muf-login-backdrop {
        background-size: 100% auto;
        background-position: center top;
        transform: none;
    }

    .muf-login-overlay {
        background: linear-gradient(180deg, rgba(8, 19, 43, 0.18) 0%, rgba(8, 19, 43, 0.5) 100%);
    }

    .muf-login-stage {
        align-items: flex-end;
        justify-content: center;
        padding: 1rem;
    }

    .muf-login-card {
        width: 100%;
        padding: 1.55rem;
        border-radius: 1.25rem;
    }

    .muf-login-card-header {
        padding-bottom: 1.1rem;
    }

    .muf-login-logo {
        width: 7.8rem;
    }

    .muf-context-pill {
        max-width: 8.5rem;
        font-size: 0.66rem;
    }

    .muf-login-copy {
        padding-top: 1.25rem;
    }
}

@media (max-height: 660px) {
    .muf-login-stage {
        align-items: flex-start;
    }
}

@media (prefers-reduced-motion: reduce) {
    .muf-microsoft-button {
        transition: none;
    }
}

/*
 * "Not you? Clear this email from this browser."
 *
 * A button, not a link, because it POSTS - it changes state on the server, and a
 * GET that deletes something is the wrong shape. Styled as a link because that is
 * what it reads as, and given a real 44px target because this site's rule is 44
 * and the last two inline controls added here were 16px and 19px.
 */
.muf-login-forget { margin: 0 0 10px; }

.muf-login-link-button {
    background: none;
    border: 0;
    color: #0072ed;
    cursor: pointer;
    display: inline-flex;
    align-items: center;
    font: inherit;
    font-size: 14px;
    min-height: 44px;
    padding: 8px 2px;
    text-decoration: underline;
}

.muf-login-link-button:hover,
.muf-login-link-button:focus { color: #004f9f; }

/*
 * The card footer now carries two lines: the photo credit the owner asked for on
 * 13 Oct 2026, and the Entra note. Stacked rather than run together, because a
 * credit and a security note are different kinds of statement and reading them as
 * one sentence makes neither land.
 */
.muf-login-card-footer { display: grid; gap: 4px; }

.muf-login-credit { color: #6b7589; }
