As 1) user is unlikely to store that on their end. 2) nowadays with lambdas isn't an easy user-side conversion.
Then we limit explosion of an already messy API.
Didn't convert Combo() "const char* items_separated_by_zeros" as this doesn't translate well when used via a ImStrv constructor, may actually aim to obsolete.
Fixed for shadowing local variable warning on 2025/06/25.
It's widely incomplete and slow, requires a printf function taking non-zero-terminated format string to work.
Might do it eventually but it's much less a problem than labels. Format string are more frequently inlined in code and tend to be small, so existing solutions at call site can work better for now.
Squashed commits (initially a commit from rokups + many rework by ocornut. keeping them separate commits made rebasing unnecessarily tricking so merged from 2024/02)
ImStrv: many fixes (see details), added imconfig class extension example, added natvis description.
ImStrv: rework toward ensuring End is always set to constant can be compile time calculated
ImStrv: using length(), fix ambiguous empty() function, fix altered behaviors, removed unused operators.
ImStrv: various tweaks and fixes. removed ImGuiTextRange from ImGuiTextFilter, fix test engine hooks, removed constructor only used twice.
nb: ImGuiSelectableFlags_NoHoldingActiveID is not used anymore. Would remove remove once we remove the unnecessary call to Selectable() from MenuItem().
Added inline redirection functions when IMGUI_DISABLE_OBSOLETE_FUNCTIONS is off.
Marked the old functions are =delete when IMGUI_DISABLE_OBSOLETE_FUNCTIONS is on, to allow for better type-checking.
The aim is to be able to use/add flags to more ImDrawList functions, and making existing functions consistents was deemed very desirable.
Because BeginMultiSelect() does `ms->ScopeRectMin = window->DC.CursorMaxPos = window->DC.CursorPos` at a time where CursorPos is already past MaxPos.y because of ItemSpacing.y. Accumulate spacing.
- IsFocused scan nav focus route.
- When covering multiple windows, draw in front most ones (grabbed FindFrontMostVisibleChildWindow() from docking branch).
While not currently a problem, it would be if ImRect::Overlaps() is changed to use <= instead of < (cc: #3976 which deal with Contains but sort of similar topic).
Changing ImDrawList::PushClipRect()'s intersect_with_current_clip_rect path to use ClipRectFull() would also fix this, but it may ambiguous there which behavior would be correct.
Amend 1ec464eb9.
SetNextItemSelectionUserData() could use g.NextItemData.SelectionUserData if we could guarantee if would be valid when nesting multi-select blocks.
But it doesn't make much of a difference as whole ImGuiMultiSelectTempData fits in 2 cache lines.
Because BoxSelectRectPrev and BoxSelectRectCurr were clamped to scope boundaries, dragging mouse outside of the scope would usually keep one axis unchanged.
Amend 15391762dd
