Version 1.92.6

This commit is contained in:
ocornut
2026-02-17 17:45:07 +01:00
parent f921abe7e7
commit e118ea06e0
9 changed files with 95 additions and 87 deletions

View File

@@ -36,35 +36,24 @@ HOW TO UPDATE?
- Please report any issue!
-----------------------------------------------------------------------
VERSION 1.92.6 WIP (In Progress)
VERSION 1.92.6 (2026-02-17)
-----------------------------------------------------------------------
Decorated log and release notes: https://github.com/ocornut/imgui/releases/tag/v1.92.6
Breaking Changes:
- Commented out legacy names obsoleted in 1.90 (Sept 2023):
- BeginChildFrame() --> BeginChild() with ImGuiChildFlags_FrameStyle flag.
- EndChildFrame() --> EndChild().
- ShowStackToolWindow() --> ShowIDStackToolWindow().
- IM_OFFSETOF() --> offsetof().
- IM_FLOOR() --> IM_TRUNC() [internal, for positive values only]
- Hashing: handling of "###" operator to reset to seed within a string identifier
doesn't include the "###" characters in the output hash anymore:
Before: GetID("Hello###World") == GetID("###World") != GetID("World");
Now: GetID("Hello###World") == GetID("###World") == GetID("World");
- This has the property of facilitating concatenating and manipulating
identifiers using "###", and will allow fixing other dangling issues.
- This will invalidate hashes (stored in .ini data) for Tables and Windows
that are using the "###" operators. (#713, #1698)
- Renamed helper macro IM_ARRAYSIZE() -> IM_COUNTOF(). Kept redirection/legacy name.
- Fonts:
- AddFontDefault() now automatically selects an embedded font between:
- AddFontDefaultVector(): new scalable font. Recommended at any higher size.
- AddFontDefaultBitmap(): classic pixel-clean font. Recommended at Size 13px with no scaling.
- AddFontDefaultVector(): new scalable font. Recommended at any higher size.
- The default selection is based on (style.FontSizeBase * FontScaleMain * FontScaleDpi)
reaching a small threshold. Prefer calling either based on your own logic.
And you can call AddFontDefaultBitmap() to ensure legacy behavior.
- Fixed handling of `ImFontConfig::FontDataOwnedByAtlas = false` which
did erroneously make a copy of the font data, essentially defeating the purpose
reaching a small threshold, but old codebases may not set any of them properly.
As as a result, it is likely that old codebase may still default to AddFontDefaultBitmap().
- Prefer explicitly calling either of them based on your own logic!
You can call AddFontDefaultBitmap() to ensure legacy behavior.
- Fixed handling of `ImFontConfig::FontDataOwnedByAtlas = false` which did
erroneously make a copy of the font data, essentially defeating the purpose
of this flag and wasting memory (undetected since July 2015 and now spotted
by @TellowKrinkle, this is perhaps the oldest bug in Dear ImGui history,
albeit for a rarely used feature!) (#9086, #8465)
@@ -74,13 +63,8 @@ Breaking Changes:
until a shutdown of the owning context or font atlas.
- The fact that handling of `FontDataOwnedByAtlas = false` was broken bypassed
the issue altogether.
- Fixed a crash when trying to use AddFont() with MergeMode=true on a font that
has already been rendered. (#9162) [@ocornut, @cyfewlp]
- Removed ImFontConfig::PixelSnapV added in 1.92 which turns out is unnecessary
(and misdocumented). Post-rescale GlyphOffset is always rounded.
- Fixed an issue where using PushFont() from the implicit/fallback "Debug" window when
its recorded state is collapsed would incorrectly early out. This would break e.g. using
direct draw-list calls such as GetForegroundDrawList() with current font. (#9210, #8865)
- Popups: changed compile-time 'ImGuiPopupFlags popup_flags = 1' default value to be '= 0' for
BeginPopupContextItem(), BeginPopupContextWindow(), BeginPopupContextVoid(), OpenPopupOnItemClick().
The default value has same meaning before and after. (#9157, #9146)
@@ -91,8 +75,8 @@ Breaking Changes:
- We have now changed this behavior to: cleanup a very old API quirk, facilitate use by
bindings, and to remove the last and error-prone non-zero default value. Also because we
deemed it extremely rare to use those helper functions with the Left mouse button!
As using the LMB would generally be triggered via another widget, e.g. a Button() +
a OpenPopup()/BeginPopup() call.
As using the LMB would generally be triggered via another widget,
e.g. a Button() + a OpenPopup()/BeginPopup() call.
- Before: The default = 1 means ImGuiPopupFlags_MouseButtonRight.
Explicitly passing a literal 0 means ImGuiPopupFlags_MouseButtonLeft.
- After: The default = 0 means ImGuiPopupFlags_MouseButtonRight.
@@ -108,6 +92,21 @@ Breaking Changes:
- BeginPopupContextItem("foo", 1); // Behavior unchanged (as a courtesy we legacy interpret 1 as ImGuiPopupFlags_MouseButtonRight, will assert if disabling legacy behaviors.
- BeginPopupContextItem("foo", 0); // !! Behavior changed !! Was Left button. Now will defaults to Right button! --> Use ImGuiPopupFlags_MouseButtonLeft.
- BeginPopupContextItem("foo", ImGuiPopupFlags_NoReopen); // !! Behavior changed !! Was Left button + flags. Now will defaults to Right button! --> Use ImGuiPopupFlags_MouseButtonLeft | xxx.
- Commented out legacy names obsoleted in 1.90 (Sept 2023):
- BeginChildFrame() --> BeginChild() with ImGuiChildFlags_FrameStyle flag.
- EndChildFrame() --> EndChild().
- ShowStackToolWindow() --> ShowIDStackToolWindow().
- IM_OFFSETOF() --> offsetof().
- IM_FLOOR() --> IM_TRUNC() [internal, for positive values only]
- Hashing: handling of "###" operator to reset to seed within a string identifier
doesn't include the "###" characters in the output hash anymore:
Before: `GetID("Hello###World") == GetID("###World") != GetID("World")`
After: `GetID("Hello###World") == GetID("###World") == GetID("World")`
- This has the property of facilitating concatenating and manipulating
identifiers using "###", and will allow fixing other dangling issues.
- This will invalidate hashes (stored in .ini data) for Tables and Windows
that are using the "###" operators. (#713, #1698)
- Renamed helper macro IM_ARRAYSIZE() -> IM_COUNTOF(). Kept redirection/legacy name.
- Backends:
- Vulkan: optional ImGui_ImplVulkanH_DestroyWindow() helper used by our example
code does not call vkDestroySurfaceKHR(): because surface is created by caller
@@ -116,7 +115,7 @@ Breaking Changes:
Other Changes:
- Fonts:
- Added AddFontDefaultVector(): a new embedded monospace scalable font: ProggyForever!
- Added `AddFontDefaultVector()`: a new embedded monospace scalable font: ProggyForever!
From https://github.com/ocornut/proggyforever:
"ProggyForever is an MIT-licensed partial reimplementation of the ProggyVector
font (originally by Tristan Grimmer), which itself is a vector-based
@@ -130,27 +129,33 @@ Other Changes:
data is ~14 KB. Embedding a scalable default font ensures that Dear ImGui can
be easily and readily used in all contexts, even without file system access.
- As always you can opt-out of the embedded font data if desired.
- AddFontDefault() now automatically selects an embedded font between
- `AddFontDefault()` now automatically selects an embedded font between
the classic pixel-looking one and the new scalable one.
Prefer calling AddFontDefaultVector() or AddFontDefaultBitmap() explicitely.
- Fixed an issue related to EllipsisChar handling, while changing
Prefer calling `AddFontDefaultVector()` or `AddFontDefaultBitmap()` explicitely.
- Fixed a crash when trying to use `AddFont()` with `MergeMode==true` on a font that
has already been rendered. (#9162) [@ocornut, @cyfewlp]
- Fixed an issue where using `PushFont()` from the implicit/fallback "Debug" window
when its recorded state is collapsed would incorrectly early out. This would break
e.g. using direct draw-list calls such as GetForegroundDrawList() with current font.
(#9210, #8865)
- Fixed an issue related to `EllipsisChar` handling, while changing
font loader or font loader flags dynamically in Style->Fonts menus.
- imgui_freetype: fixed overwriting ImFontConfig::PixelSnapH when hinting
- imgui_freetype: fixed overwriting `ImFontConfig::PixelSnapH` when hinting
is enabled, creating side-effects when later disabling hinting or
dynamically switching to stb_truetype rasterizer.
- Post rescale GlyphOffset is always rounded.
- Post rescale `ImFontConfig::GlyphOffset` is always rounded.
- Adding new fonts after removing all fonts mid-frame properly updates current state.
- Textures:
- Fixed a building issue when ImTextureID is defined as a struct.
- Fixed a building issue when `ImTextureID` is defined as a struct.
- Fixed displaying texture # in Metrics/Debugger window.
- Menus:
- Fixed MenuItem() label position and BeginMenu() arrow/icon/popup positions,
- Fixed `MenuItem()` label position and `BeginMenu()` arrow/icon/popup positions,
when used inside a line with a baseline offset.
- Made navigation into menu-bar auto wrap on X axis. (#9178)
- TreeNode:
- Fixed highlight position when used inside a line with a large text baseline offset.
(never quite worked in this situation; but then most of the time the text
baseline offset ends up being zero or FramePadding.y for a given line).
baseline offset ends up being zero or `FramePadding.y` for a given line).
- Tables:
- Fixed an issue where a very thin scrolling table would advance parent layout
slightly differently depending on its visibility (caused by a mismatch
@@ -161,70 +166,71 @@ Other Changes:
data has missing or duplicate values. (#9108, #4046)
- ColorEdit:
- Added R/G/B/A color markers next to each component (enabled by default).
- Added ImGuiColorEditFlags_NoColorMarkers to disable them.
- Added style.ColorMarkerSize to configure width of color component markers.
- Added `ImGuiColorEditFlags_NoColorMarkers` to disable them.
- Added `style.ColorMarkerSize` to configure width of color component markers.
- Sliders, Drags:
- Added ImGuiSliderFlags_ColorMarkers to opt-in adding R/G/B/A color markers
- Added `ImGuiSliderFlags_ColorMarkers` to opt-in adding R/G/B/A color markers
next to each components, in multi-components functions.
- Added a way to select a specific marker color.
- InputText:
- InputTextMultiline(): fixed a minor bug where Shift+Wheel would allow a small
horizontal scroll offset when there should be none. (#9249)
- ImGuiInputTextCallbackData: SelectAll() also sets CursorPos to SelectionEnd.
- ImGuiInputTextCallbackData: Added SetSelection() helper.
- ImGuiInputTextCallbackData: Added ID and EventActive helpers. (#9174)
- ImGuiInputTextCallbackData: `SelectAll()` also sets `CursorPos` to `SelectionEnd`.
- ImGuiInputTextCallbackData: Added `SetSelection()` helper.
- ImGuiInputTextCallbackData: Added `ID` and `EventActivated` members. (#9174)
- Text, InputText:
- Reworked word-wrapping logic:
- Try to not wrap in the middle of contiguous punctuations. (#8139, #8439, #9094)
- Try to not wrap between a punctuation and a digit. (#8503)
- Inside InputTextMultiline() with _WordWrap: prefer keeping blanks at the
end of a line rather than at the beginning of next line. (#8990, #3237)
- Fixed low-level word-wrapping function reading from *text_end when passed
- Inside `InputTextMultiline()` with WordWrap enabled: prefer keeping blanks at
the end of a line rather than at the beginning of next line. (#8990, #3237)
- Fixed low-level word-wrapping function reading from `*text_end` when passed
a string range. (#9107) [@achabense]
- Changed RenderTextEllipsis() logic to not trim trailing blanks before
- Changed `RenderTextEllipsis()` logic to not trim trailing blanks before
the ellipsis, making ellipsis position more consistent and not arbitrary
hiding the possibility of multiple blanks. (#9229)
- Nav:
- Fixed remote/shortcut InputText() not teleporting mouse cursor when
nav cursor is visible and `io.ConfigNavMoveSetMousePos` is enabled.
- Fixed a looping/wrapping issue when done in menu layer. (#9178)
- Fixed a looping/wrapping issue when used in menu layer. (#9178)
- Fixed speed scale for resizing/moving with keyboard/gamepad. We incorrectly
used io.DisplayFramebufferScale (very old code), effectively making those
actions faster on macOS/iOS retina screens.
used `io.DisplayFramebufferScale` as a scaling factor (very old code),
effectively making those actions faster on macOS/iOS retina screens.
(changed this to use a style scale factor that's not fully formalized yet)
- Fixed an UBSan warning when using in a ListClipper region . (#9160)
- Fixed an UBSan warning when using in a `ImGuiListClipper` region . (#9160)
- Scrollbar: fixed a codepath leading to a divide-by-zero (which would not be
noticeable by user but detected by sanitizers). (#9089) [@judicaelclair]
- InvisibleButton: allow calling with size (0,0) to fit to available content
size. (#9166, #7623)
- Tooltips, Disabled: fixed EndDisabledOverrideReenable() assertion when
- Tooltips, Disabled: fixed `EndDisabledOverrideReenable()` assertion when
nesting a tooltip in a disabled block. (#9180, #7640) [@RegimantasSimkus]
- Added GetItemFlags() in public API for consistency and to expose generic
- Added `GetItemFlags()` in public API for consistency and to expose generic
flags of last submitted item. (#9127)
- Log/Capture: fixed erroneously injecting extra carriage returns in output
buffer when ItemSpacing.y > FramePadding.y + 1.
- Misc: fixed build on ARM64/ARM64EC targets trying to use SSE/immintrin.h.
(#9209, #5943, #4091) [@navvyswethgraphics]
- Log/Capture:
- Fixed erroneously injecting extra carriage returns in output text buffer
when `ItemSpacing.y` > `FramePadding.y + 1` while emitting items.
- Images:
- Added style.ImageRounding, ImGuiStyleVar_ImageRounding to configure
rounding of Image() widgets. (#2942, #845)
- ImageButton() doesn't use a clamped style.FrameRounding value but instead
adjust inner image rounding when FramePadding > FrameRounding. (#2942, #845)
- Added `style.ImageRounding`, `ImGuiStyleVar_ImageRounding `to configure
rounding of `Image()` widgets. (#2942, #845)
- `ImageButton()` doesn't use a clamped `style.FrameRounding` value but instead
adjust inner image rounding when `FramePadding > `FrameRounding`. (#2942, #845)
- Shortcuts:
- IsItemHovered() without ImGuiHoveredFlags_AllowWhenBlockedByActiveItem
- IsItemHovered() without `ImGuiHoveredFlags_AllowWhenBlockedByActiveItem`
doesn't filter out the signal when activated item is a shortcut remote activation;
(which mimicks what's done internally in the ItemHoverable() function). (#9138)
(which mimicks what's done internally in the `ItemHoverable()` function). (#9138)
- Fixed tooltip placement being affected for a frame when located over an item
activated by SetNextItemShortcut(). (#9138)
activated by `SetNextItemShortcut()`. (#9138)
- Error Handling:
- Improve error handling and recovery for EndMenu()/EndCombo(). (#1651, #9165, #8499)
- Improve error handling and recovery for TableSetupColumn().
- Improved error handling and recovery for `EndMenu()`/`EndCombo()`. (#1651, #9165, #8499)
- Improved error handling and recovery for `TableSetupColumn()`.
- Debug Tools:
- Debug Log: fixed incorrectly printing characters in IO log when submitting
non-ASCII values to `io.AddInputCharacter()`. (#9099)
- Debug Log: can output to debugger on Windows. (#5855)
- Debug Log: can output to debugger on Windows via Win32 `OutputDebugString()` (#5855)
- Demo:
- Slightly improve Selectable() demos. (#9193)
- Slightly improve `Selectable()` demos. (#9193)
- Backends:
- DirectX10: added `SamplerNearest` in `ImGui_ImplDX10_RenderState`.
(+renamed `SamplerDefault` to `SamplerLinear`, which was tagged as beta API)
@@ -232,7 +238,7 @@ Other Changes:
(+renamed `SamplerDefault` to `SamplerLinear`, which was tagged as beta API)
- GLFW: Avoid repeated `glfwSetCursor()` / `glfwSetInputMode()` unnecessary calls.
Lowers overhead for very high framerates (e.g. 10k+ FPS). [@maxliani]
- GLFW: Added IMGUI_IMPL_GLFW_DISABLE_X11 / IMGUI_IMPL_GLFW_DISABLE_WAYLAND to
- GLFW: Added `IMGUI_IMPL_GLFW_DISABLE_X11` / `IMGUI_IMPL_GLFW_DISABLE_WAYLAND` to
forcefully disable either. (#9109, #9116)
Try to set them automatically if headers are not accessible. (#9225)
- OpenGL3: Fixed embedded loader multiple init/shutdown cycles broken on some
@@ -244,14 +250,14 @@ Other Changes:
but a better default for X11 users). Waiting for a drag to start mouse capture leads to
input drops when dragging after clicking on the edge of a window.
(#3650, #6410, #9235, #3956, #3835)
- SDL2, SDL3: added ImGui_ImplSDL2_SetMouseCaptureMode()/ImGui_ImplSDL3_SetMouseCaptureMode()
- SDL2, SDL3: added `ImGui_ImplSDL2_SetMouseCaptureMode()`/`ImGui_ImplSDL3_SetMouseCaptureMode()`
function for X11 users to disable mouse capturing/grabbing. (#3650, #6410, #9235, #3956, #3835)
- When attached to a debugger may want to call:
- `ImGui_ImplSDL3_SetMouseCaptureMode(ImGui_ImplSDL3_MouseCaptureMode_Disabled);`
- But you can also configure your system or debugger to automatically release
mouse grab when crashing/breaking in debugger, e.g.
- console: `setxkbmap -option grab:break_actions && xdotool key XF86Ungrab`
- or use a GDB script to capture SDL_CaptureMouse(false). See #3650.
- or use a GDB script to call SDL_CaptureMouse(false). See #3650.
- On platforms other than X11 this is unnecessary.
- SDL_GPU3: added `SamplerNearest` in `ImGui_ImplSDLGPU3_RenderState`.
- SDL_GPU3: macOS version can use MSL shaders in order to support macOS 10.14+
@@ -261,7 +267,7 @@ Other Changes:
selects `VkSwapchainCreateInfoKHR`'s `compositeAlpha` value based on
`cap.supportedCompositeAlpha`, which seems to be required on some Android
devices. (#8784) [@FelixStach]
- WebGPU: fixes for Emscripten 5.0.0 (note: our examples currently don't build with 5.0.1).
- WebGPU: fixes for Emscripten 5.0.0 (note: current examples do not build with 5.0.1).
- Win32: handle `WM_IME_CHAR`/`WM_IME_COMPOSITION` to support Unicode inputs on
MBCS (non-Unicode) Windows. (#9099, #3653, #5961) [@ulhc, @ocornut, @Othereum]
- Win32: minor optimization not submitting gamepad input if packet number has not