fix(ui): fellowship checkbox count, row-text + ListBox Flush doc corrections, promote probe findings to assertions
Mechanism SHOULD-FIX 4: SocialFellowshipPageController's class doc said 0x1000026B holds "all three visible option checkboxes" — the fixture and the decomp both authors FOUR (Ignore Fellowship Requests / Auto-Accept Requests / Share XP / Share Loot, 0x10000270-0x10000273). Corrected. Mechanism SHOULD-FIX 9: SocialPanelRowText.FindDeepest's doc promised "the deepest UiText descendant" but the implementation returns the LAST match in pre-order traversal order, which only equals the deepest when the subtree is a single chain. Both real row templates ARE single chains today, so behavior is unaffected — the doc now describes what the code actually does instead of a stronger guarantee it doesn't implement. Blast SHOULD-FIX 5: UiTemplateListBox.Flush()'s doc only mentioned the ContentHeight reset; UiScrollablePanel.ClearContent() also resets scroll position to 0, which the sibling UiItemList.Flush() (same method name, different semantics) does NOT do. Documented explicitly, including the UX cost this creates for a scrolled-in Friends/Squelch roster once its scrollbar is wired (this fix round's blast MF-1) — flagged for whoever revisits Friends/Squelch scrolling next rather than silently fixed as an unasked behavior change. Mechanism SHOULD-FIX 3: SocialPanelLiveMountProbeTests printed two headline findings (the 0x10000492-authored-twice count, page exclusivity after ActivateTabBehavior) without ever asserting them — a future importer regression collapsing/dropping an instance, or breaking exclusivity, could not fail this test. Both are now real assertions (Assert.Equal(2, passupCount); exactly one page Visible and it is Allegiance). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
parent
35c40a9b56
commit
ae77270939
4 changed files with 56 additions and 6 deletions
|
|
@ -16,9 +16,20 @@ namespace AcDream.App.UI.Layout;
|
|||
/// </summary>
|
||||
internal static class SocialPanelRowText
|
||||
{
|
||||
/// <summary>Returns the deepest <see cref="UiText"/> descendant of
|
||||
/// <paramref name="row"/> (pre-order, last match wins), or null if the
|
||||
/// built row contains no text element at all.</summary>
|
||||
/// <summary>
|
||||
/// Returns the LAST <see cref="UiText"/> found by a pre-order walk of
|
||||
/// <paramref name="row"/> (row itself, then each child's own subtree in
|
||||
/// order), or null if the built row contains no text element at all.
|
||||
/// <b>Fix-round mechanism SF-9:</b> this equals the truly DEEPEST match
|
||||
/// only when the subtree is a single chain — both real row templates
|
||||
/// (Friends <c>0x10000519</c>, Squelch <c>0x10000541</c>) ARE single
|
||||
/// chains (row root -> one Type-0xC text child -> one more Type-0xC
|
||||
/// text grandchild, per the FA3 live-mount probe), so this method's
|
||||
/// actual "last match in traversal order" behavior and the class doc's
|
||||
/// "deepest" framing agree for both templates in production today. A
|
||||
/// row template with a BRANCHING subtree (multiple text leaves at
|
||||
/// different depths) would expose the difference.
|
||||
/// </summary>
|
||||
public static UiText? FindDeepest(UiElement row)
|
||||
{
|
||||
UiText? found = row as UiText;
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue