Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
Deploy AIOS web (React glide grid + FastAPI slice)
Browse files- RELEASES.json +1 -1
- VERSION +1 -1
- web/src/customer-grid/OverlaySurface.tsx +5 -1
- web/src/index.css +30 -0
RELEASES.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"current": "
|
| 3 |
"releases": [
|
| 4 |
{
|
| 5 |
"version": "v29",
|
|
|
|
| 1 |
{
|
| 2 |
+
"current": "c05da70",
|
| 3 |
"releases": [
|
| 4 |
{
|
| 5 |
"version": "v29",
|
VERSION
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
|
|
|
|
| 1 |
+
c05da70
|
web/src/customer-grid/OverlaySurface.tsx
CHANGED
|
@@ -341,7 +341,11 @@ export function AnchoredOverlay({
|
|
| 341 |
<div
|
| 342 |
ref={panelRef}
|
| 343 |
id={id}
|
| 344 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 345 |
style={style}
|
| 346 |
role={role}
|
| 347 |
aria-label={ariaLabel}
|
|
|
|
| 341 |
<div
|
| 342 |
ref={panelRef}
|
| 343 |
id={id}
|
| 344 |
+
// `cg-overlay` FIRST and unconditionally: this node is a direct child of <body>
|
| 345 |
+
// (see `BodyPortal`), so it competes with the app's dialogs in the root stacking
|
| 346 |
+
// context and needs a z-index that says so. The caller's own class still decides
|
| 347 |
+
// everything else about it. See `body > .cg-overlay` in index.css.
|
| 348 |
+
className={"cg-overlay" + (className ? ` ${className}` : "")}
|
| 349 |
style={style}
|
| 350 |
role={role}
|
| 351 |
aria-label={ariaLabel}
|
web/src/index.css
CHANGED
|
@@ -57,6 +57,23 @@
|
|
| 57 |
--cg-z-menu: 50;
|
| 58 |
--cg-z-column: 60;
|
| 59 |
--cg-z-drawer: 70;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
/* --- Loopable C1 tokens (wave-8 I7c) ------------------------------------
|
| 62 |
ADDITIVE on purpose. The `--cg-*` tokens above are NOT redefined because
|
|
@@ -1238,6 +1255,19 @@ body {
|
|
| 1238 |
position: relative;
|
| 1239 |
display: inline-flex;
|
| 1240 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1241 |
.cg-pop {
|
| 1242 |
z-index: var(--cg-z-menu);
|
| 1243 |
min-width: min(220px, calc(100vw - 16px));
|
|
|
|
| 57 |
--cg-z-menu: 50;
|
| 58 |
--cg-z-column: 60;
|
| 59 |
--cg-z-drawer: 70;
|
| 60 |
+
/* ββ THE PORTAL'D OVERLAY STEP, AND IT IS ABOVE THE DIALOG STEP ON PURPOSE.
|
| 61 |
+
`AnchoredOverlay` renders through `BodyPortal` β a DIRECT CHILD OF <body>, so it
|
| 62 |
+
resolves in the ROOT stacking context and its z-index competes with the app's
|
| 63 |
+
dialogs rather than with the surface it was opened from. `.cg-fsel-pop` carries
|
| 64 |
+
`.cg-pop` (50) and `.set-backdrop` is 70, so every field picker opened inside
|
| 65 |
+
Manage user PAINTED UNDERNEATH THE SETTINGS DIALOG: open, visible, 232x263px,
|
| 66 |
+
and behind. `document.elementFromPoint` at the popover's own centre returned
|
| 67 |
+
`.cg-cond-row` β an element of the dialog β so every click landed on the dialog.
|
| 68 |
+
MEASURED on the deployed build, not deduced.
|
| 69 |
+
β WHY THE OP DROPDOWN NEXT TO IT WORKED AND HID THIS FOR SO LONG: it is a native
|
| 70 |
+
`<select>`, which the browser paints in the TOP LAYER whatever the z-index says.
|
| 71 |
+
One condition row, two controls, one of them exempt from the bug by being a
|
| 72 |
+
browser widget. The reader concludes "the filter works, that one field is stuck".
|
| 73 |
+
β ABOVE 80 (`.shell-newdb`'s scrim) and 90 (`.shell-nav-tip`) too: a transient
|
| 74 |
+
panel anchored to a control the user just clicked belongs above every surface,
|
| 75 |
+
because it is only on screen while they are using it. */
|
| 76 |
+
--cg-z-overlay: 95;
|
| 77 |
|
| 78 |
/* --- Loopable C1 tokens (wave-8 I7c) ------------------------------------
|
| 79 |
ADDITIVE on purpose. The `--cg-*` tokens above are NOT redefined because
|
|
|
|
| 1255 |
position: relative;
|
| 1256 |
display: inline-flex;
|
| 1257 |
}
|
| 1258 |
+
/* β EVERY `AnchoredOverlay`, not `.cg-fsel-pop` alone. The class is stamped by
|
| 1259 |
+
`OverlaySurface` on the portal'd node itself, so a panel added tomorrow inherits this
|
| 1260 |
+
instead of re-finding the bug. `body > ` rather than a bare class is doing real work: it
|
| 1261 |
+
raises SPECIFICITY to (0,1,1) so it beats `.cg-pop`'s own `z-index: var(--cg-z-menu)` at
|
| 1262 |
+
(0,1,0) regardless of source order, and it cannot match the INLINE `.cg-pop.set-perm-pop`
|
| 1263 |
+
panels mounted inside the permission editor β those are not children of <body>, they must
|
| 1264 |
+
STAY at 50, and `.set-confirm-wrap` (70) is what covers them when Save asks its question.
|
| 1265 |
+
Raising `.cg-pop` wholesale would put the filter panel back over that confirm dialog, which
|
| 1266 |
+
is a bug this file already records fixing. */
|
| 1267 |
+
body > .cg-overlay {
|
| 1268 |
+
z-index: var(--cg-z-overlay);
|
| 1269 |
+
}
|
| 1270 |
+
|
| 1271 |
.cg-pop {
|
| 1272 |
z-index: var(--cg-z-menu);
|
| 1273 |
min-width: min(220px, calc(100vw - 16px));
|