Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[emscripten + glfw]: Added support for GLFW3 contrib port #7647

Closed
wants to merge 6 commits into from

Conversation

ypujante
Copy link
Contributor

@ypujante ypujante commented Jun 2, 2024

This is the PR implementing the changes as discussed in PR #7520. Due to the changes + merge conflicts, I decided to start clean and do a brand new PR.

As discussed, this PR adds support for the GLFW3 contrib port, but does NOT force the user to switch to it.

  • I have not changed Makefile.emscripten in example_glfw_wgpu and example_glfw_opengl3 and as a result, building using these Makefiles, will build the same code as before
  • I have changed CMakeLists.txt for example_glfw_wgpu to automatically detect the version of emscripten used and if it can use the new port, it uses it: this gives an example on how to build with the new port
  • As discussed I also took the opportunity to fix Emscripten build runtime error: Unable to preventDefault inside passive event listener #7600 (using the canvas selector as opposed to the document addresses the issue)

The most notable changes visible when using the GLFW3 contrib port are (which can be tested with the live demos)

  • the gamepad is now working properly
  • copy to clipboard works (Menu: Tools / About Dear ImGui / Config/Build Configuration / Copy to clipboard)
  • Hi DPI by default

Live demos:

  • Build using -sUSE_GLFW=3: demo
  • Build using --use-port=contrib.glfw3: demo

- implements GFLW 3.4.0 features + bug fixes (like full gamepad support)
- fixes "Emscripten build runtime error: Unable to preventDefault inside passive event listener"
@ocornut
Copy link
Owner

ocornut commented Jun 3, 2024

Thank you for updating this.

EMSCRIPTEN_USE_GLFW3 seems a little confusingly named. Maybe I would use a more explicit name:
EMSCRIPTEN_USE_PORT_CONTRIB_GLFW3 vs EMSCRIPTEN_USE_EMBEDDED_GLFW3 ?

I'll read the rest when I have some time.

@ypujante
Copy link
Contributor Author

ypujante commented Jun 3, 2024

I agree it is a bit confusing. I renamed it per your suggestion.

@ocornut
Copy link
Owner

ocornut commented Jun 10, 2024

  • Why is emscripten_set_wheel_callback called in the ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() function?
  • I think the printf("") may be replaced by an assert() with comments.

@ypujante
Copy link
Contributor Author

  • Why is emscripten_set_wheel_callback called in the ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() function?

You asked me to fix the #7600 issue for the embedded path and the fix is to call emscripten_set_wheel_callback with the selector. The prior code was calling emscripten_set_wheel_callback with a static value (for the document) and was being called in ImGui_ImplGlfw_Init. There is no access to the selector (in the embedded path) until you call ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() while providing it as a parameter. So that is the reason why I had to move it here.

* I think the `printf("")` may be replaced by an assert() with comments.

Ok I will change it.

@ocornut
Copy link
Owner

ocornut commented Jun 10, 2024

Thanks! I didn't realize this was the fix for #7600. I'll try to test this now.

@ocornut
Copy link
Owner

ocornut commented Jun 10, 2024

I wonder if we should:

  • rename ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() to ImGui_ImplGlfw_InstallEmscriptenCallbacks()
  • we can easily add an inline redirect under #ifndef IMGUI_DISABLE_OBSOLETE_FUNCTIONS block.
  • KEEP calling ImGui_ImplGlfw_InstallEmscriptenCallbacks() even for your port. We disable the code inside imgui_impl_glfw.cpp but it seems sane/safe to request users to call this so we have a backup path if ever some function in your port requires extra code?

What is the "window" string passed to emscripten_glfw_make_canvas_resizable() ? is that part of the template?

Would it make some sense to ask both window and canvas names in ImGui_ImplGlfw_InstallEmscriptenCallbacks() and do that automatically?

@ypujante
Copy link
Contributor Author

I made the change you requested printf -> assert.

I made it as a static_assert because the runtime assert was pretty horrible from a user perspective

Screenshot 2024-06-10 at 05 59 18

At least with the static_assert the error message is at compilation time and add some detail:

../../backends/imgui_impl_glfw.cpp:867:19: error: static assertion failed: [ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback] When using --use-port=contrib.glfw3, you should include <GLFW/emscripten_glfw3.h> and call emscripten_glfw_make_canvas_resizable instead
  867 |     static_assert(false, "[ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback] When using --use-port=contrib.glfw3, you should include <GLFW/emscripten_glfw3.h> and call emscripten_glfw_make_canvas_resizable instead");
      |                   ^~~~~

In regards to your last questions:

  • rename ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() to ImGui_ImplGlfw_InstallEmscriptenCallbacks()
  • we can easily add an inline redirect under #ifndef IMGUI_DISABLE_OBSOLETE_FUNCTIONS block.
  • KEEP calling ImGui_ImplGlfw_InstallEmscriptenCallbacks() even for your port. We disable the code inside imgui_impl_glfw.cpp but it seems sane/safe to request users to call this so we have a backup path if ever some function in your port requires extra code?

I don't think I have an opinion one way or another. I don't envision the port to ever require something that would need to be added in ImGui_ImplGlfw_InstallEmscriptenCallbacks because it would mean that every other (non ImGui) projects my port is used would require this callbacks to be installed as well and I don't think that would be good for the end user.

But it doesn't cost much to do it the way you suggest and I am fine with it if that is the direction you want to go with.

What is the "window" string passed to emscripten_glfw_make_canvas_resizable() ? is that part of the template?

The documentation explains in detail the parameters provided to this function which can handle 3 different use cases. The "normal" use case for ImGui is to use the full window, and this corresponds to the css selector "window". As you can see in my WIP videos on a new project I am working on, (WIP1 and WIP2), in this instance I DON'T use "window" but the canvas selector itself, thus having ImGui run in a resizable frame in the browser.

Would it make some sense to ask both window and canvas names in ImGui_ImplGlfw_InstallEmscriptenCallbacks() and do that automatically?

I think the issue is that, as I mentioned, with the new API you don't necessarily use the full window, so I am not sure how the user could do that. And how would you pass the 3rd parameter (which is the handle to resize the canvas when you want one)?

- added ImGui_ImplGlfw_InstallEmscriptenCallbacks and deprecated ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback
@ypujante
Copy link
Contributor Author

After spending more time thinking about your proposed changes (add ImGui_ImplGlfw_InstallEmscriptenCallbacks and deprecate ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback) I tried to implement it and I think I really like the suggestion.

It makes it easier for regular ImGui user. Although I was worried about the fact that the emscripten_glfw_make_canvas_resizable allow for 3 different use cases (that are not covered with this solution), the reality is probably 99.9% of ImGui users will not care about not being full window. For the remaining users who want a more advanced control, this solution will still work as a user can write something like this:

// main.cpp
#ifdef __EMSCRIPTEN__
#include <GLFW/emscripten_glfw3.h>
#endif
// ....
#ifdef __EMSCRIPTEN__
    ImGui_ImplGlfw_InstallEmscriptenCallbacks(window, "#canvas");  // this will make it full window
    emscripten_glfw_make_canvas_resizable(window, "#xxx", "#yyy"); // this will make it resizable via #xxx and handle #yyy
#endif

In other words, my code supports calling emscripten_glfw_make_canvas_resizable multiple times and does the right thing...

In conclusion, I would go for this solution which covers 99.9% of the use cases while still allowing to use the more advanced usage. It is less discoverable, but I am planning to write examples demonstrating it (including open sourcing the project I linked to in my previous message).

I marked ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback deprecated which then raises a nice warning during compilation:

main.cpp:108:5: warning: 'ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback' is deprecated: Use ImGui_ImplGlfw_InstallEmscriptenCallbacks instead [-Wdeprecated-declarations]
  108 |     ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback("#canvas");
      |     ^
../../backends/imgui_impl_glfw.h:35:3: note: 'ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback' has been explicitly marked deprecated here
   35 | [[deprecated("Use ImGui_ImplGlfw_InstallEmscriptenCallbacks instead")]]
      |   ^
1 warning generated.

@ypujante
Copy link
Contributor Author

Hi @ocornut, I just found out an issue which is reproducible in BOTH the old/embedded path as well as the new one on macOS

  • go to Demo/Inputs & Focus
  • Hit Meta (Super)
  • While holding Meta, Press V
  • Release V
  • Release Meta

"V" is still pressed

Meta_V

I am going to investigate of why this is happening and I will let you know what I find.

@ypujante
Copy link
Contributor Author

As I pointed out in my comment:

Since this problem existed before I don't see a reason why this should prevent this PR to be merged (unless there are other reasons to prevent the merge of course).

@ypujante
Copy link
Contributor Author

ypujante commented Jul 6, 2024

I merged all the changes (I saw there was a conflict which I fixed). I also rebuilt/republished the demo built with the latest emscripten (1.3.62) which contains the latest version of emscripten-glfw.

In particular this update contains the following changes:

  • addresses the Meta key issue on macOS as mentioned above
  • allow for copy to clipboard (so you can now copy the configuration to the clipboard)
Dear ImGui 1.91.0 WIP (19093)
--------------------------------
sizeof(size_t): 4, sizeof(ImDrawIdx): 2, sizeof(ImDrawVert): 20
define: __cplusplus=201703
define: IMGUI_DISABLE_FILE_FUNCTIONS
define: __GNUC__=4
define: __clang_version__=19.0.0git (https:/github.com/llvm/llvm-project c00ada070207979f092be9046a02fcfff8b9f9ce)
define: __EMSCRIPTEN__
--------------------------------
io.BackendPlatformName: imgui_impl_glfw
io.BackendRendererName: imgui_impl_webgpu
io.ConfigFlags: 0x00000003
 NavEnableKeyboard
 NavEnableGamepad
io.ConfigInputTextCursorBlink
io.ConfigWindowsResizeFromEdges
io.ConfigMemoryCompactTimer = 60.0
io.BackendFlags: 0x0000000E
 HasMouseCursors
 HasSetMousePos
 RendererHasVtxOffset
--------------------------------
io.Fonts: 1 fonts, Flags: 0x00000000, TexSize: 512,64
io.DisplaySize: 1232.00,688.00
io.DisplayFramebufferScale: 2.00,2.00
--------------------------------
style.WindowPadding: 8.00,8.00
style.WindowBorderSize: 1.00
style.FramePadding: 4.00,3.00
style.FrameRounding: 0.00
style.FrameBorderSize: 0.00
style.ItemSpacing: 8.00,4.00
style.ItemInnerSpacing: 4.00,4.00

@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

Is this mergeable? Topics/PR going over multiple things tends to be overwhelming/hard to process for me.
I'll just need to fix the coding style but I can do that myself. I will also likely nuke the obsolete stuff and turn them into comments, there's too much too maintain already and Emscripten is fast moving.

@ypujante
Copy link
Contributor Author

ypujante commented Jul 8, 2024

As far as I know yes, this is mergeable. Due to the fact that you have to opt-in to use the new implementation, and the old implementation is still the default, it should be fairly low risk in the gran scheme of things.

@ypujante
Copy link
Contributor Author

ypujante commented Jul 8, 2024

Topics/PR going over multiple things tends to be overwhelming/hard to process for me.

Let me try to recap it for you and summarize it. This PR includes:

  1. Introduced ability to swap the emscripten GLFW implementation from the embedded one (javascript) to the more modern/fully featured external one (cpp):
  2. Embedded implementation is still the default and can always be used. The CMakeLists.txt for examples_glfw_wgpu was changed to detect if the new implementation can be used and if so uses it (can be overridden to revert the behavior, documented in the file) .
  3. Fixed the warning from Emscripten build runtime error: Unable to preventDefault inside passive event listener #7600 in the embedded path

@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

It also breaks GLFW compilation without Emscripten. I'll finish this and split into two commits.

@ypujante
Copy link
Contributor Author

ypujante commented Jul 8, 2024

I sincerely apologize for that oversight. If you want me to address the problem I'm more than happy to look into it

@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

No worries I'll finish it, but it's always easier when PR are separated in commits. Here I'll want separate the contrib port support from the ImGui_ImplGlfw_InstallEmscriptenCallbacks() rename.

ocornut pushed a commit that referenced this pull request Jul 8, 2024
…mscriptenCanvasResizeCallback() to ImGui_ImplGlfw_InstallEmscriptenCallbacks(), added GLFWwindow* parameter. (#7647, #7600)

+ Fixed Emscripten warning when using mouse wheel on some setups.
@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

Merged as 6816789 and 2937339 (with various small changes)
I got rid of ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() redirect because we have too much cruft, Emscripten is sort of a moving target, and the fix on user's end is trivial. It is kept as a comment!
Thanks!

@ocornut ocornut closed this Jul 8, 2024
@ypujante
Copy link
Contributor Author

ypujante commented Jul 8, 2024

@ocornut Thank you for merging my changes and again, I apologize for the issues. I am not sure if I could have triggered a CI run myself but it would have helped in knowing I had broken something. I will do better next time.

I also wanted to reiterate that you should not hesitate to contact me if there is any issue with the code changes that I did and I will do my best to respond as quickly as possible. I can also help you with questions about emscripten in general (like you involved me for the "open a window" feature) and will be more than happy to help if I can.

@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

GitHub doesn’t seem to run CI without approval unless a commit(?) or pr(?) has previously been merged from the same author. But i am not sure which rule uses precisely: as I tend to amend commits sometimes github doesn’t consider a PR as merged even if it is. I could have validated the PR earlier for CI but I also tend to forget to do this before I get around to review something.

I also forgot to add the glfw contrib mention to the main changelog.txt i will do this on a future commit (i am afk now).

Thanks for your offer to help, much appreciated.

ocornut added a commit that referenced this pull request Jul 8, 2024
@ocornut
Copy link
Owner

ocornut commented Jul 8, 2024

Pushed additional fix a8e96ae (caused by my own leftover during a test)

@ypujante
Copy link
Contributor Author

@ocornut I just wanted to let you know that I did a sanity check after you made my changes to master (a8e96ae) and everything looks fine with both embedded/port backends (I tested Makefile.emscripten and CMakeLists.txt)

In case you are interested, I think there is a piece of information that could be added to the Config/Build Information of the About window: the version of emscripten used to compile (since issues could be tied to which version of emscripten is used to compile...). It is very easy to get this data:

#include <emscripten/version.h>
// ...
ImGui::Text("emscripten: %d.%d.%d\n", __EMSCRIPTEN_major__, __EMSCRIPTEN_minor__, __EMSCRIPTEN_tiny__);

This defines are described in the documentation

Ciachociech added a commit to Ciachociech/imgui that referenced this pull request Jul 24, 2024
* Version 1.90.8

* Version 1.90.9 WIP

* Internals: renamed HoveredIdDisabled to HoveredIdIsDisabled for consistency.

* Examples: GLFW+Vulkan: handle swap chain resize even without Vulkan returning VK_SUBOPTIMAL_KHR (ocornut#7671)

* Examples: SDL+Vulkan: handle swap chain resize even without Vulkan returning VK_SUBOPTIMAL_KHR (ocornut#7671)

* Removed old nested structure: renaming ImGuiStorage::ImGuiStoragePair type to ImGuiStoragePair (simpler for many languages).

* Internals: made ImLowerBound() accessible in internals + take a span. + rearrange child/popup/tooltips section.

Because upcoming rework of ImGuiSelectionBasicStorage will want to do a lower bound on a span.

* IO: do not disable io.ConfigWindowsResizeFromEdges when ImGuiBackendFlags_HasMouseCursors is not set by backend.

Amend 42bf149

* Style: (Breaking) renamed ImGuiCol_TabActive -> ImGuiCol_TabSelected, ImGuiCol_TabUnfocused -> ImGuiCol_TabDimmed, ImGuiCol_TabUnfocusedActive -> ImGuiCol_TabDimmedSelected.

Amend ocornut#261, ocornut#351

* TabBar, Style: added ImGuiTabBarFlags_DrawSelectedOverline and ImGuiCol_TabSelectedOverline, ImGuiCol_TabDimmedSelectedOverline.

* Drag and Drop: BeginDragDropSource() with ImGuiDragDropFlags_SourceExtern. (ocornut#143)

Amend 0c6e260

* Drag and Drop: Fixes an issue when elapsing payload would be based on last payload frame instead of last drag source frame.

* Internals: added ImGuiContext::ContextName optionally used by debug log and to facilitate debugging.

* Drag and Drop: comments, debug log entries.

* Drag and Drop: BeginDragDropSource() with ImGuiDragDropFlags_SourceExtern assume a mouse button being pressed. (ocornut#143)

* Drag and Drop: (Breaking) renamed ImGuiDragDropFlags_SourceAutoExpirePayload to ImGuiDragDropFlags_PayloadAutoExpire. (ocornut#1725, ocornut#143)

* Drag and Drop: Added ImGuiDragDropFlags_PayloadNoCrossContext and ImGuiDragDropFlags_PayloadNoCrossProcess flags.

* Ignore .ini file with other suffixes.

* Fixed build warning.

* IO: added ClearInputMouse(). made ClearInputKeys() not clear mouse data. (ocornut#4921)

Amend 6aa408c

* IO: added ImGuiConfigFlags_NoKeyboard for consistency and convenience. (ocornut#4921)

# Conflicts:
#	imgui.h
#	imgui_demo.cpp

* Nav: CTRL+Tab overlay display context name if any.

* Internals: storing HoveredWindowBeforeClear for use by multi-context compositor drag and drop propagation.

# Conflicts:
#	imgui.cpp
#	imgui_internal.h

* (Breaking) Move ImGuiWindowFlags_NavFlattened to ImGuiChildFlags_NavFlattened. (ocornut#7687)

* Backends: SDL3: Follow SDL3 removal of keysym field in SDL_KeyboardEvent (ocornut#7729)

* Demo: Style Editor: clarify how _CalcCircleAutoSegmentCount() doesn't always get exact final segment count. (ocornut#7731)

* Backends: Vulkan: Remove Volk/ from volk.h #include directives (ocornut#7722, ocornut#6582, ocornut#4854)

* Metrics/Debugger: Browsing a Storage perform hover lookup on identifier.

* Viewports: Backported 'void* ImGuiViewport::PlatformHandle' from docking branch for use by backends.

* Backends: SDL3: Update for SDL_StartTextInput()/SDL_StopTextInput() API changes. (ocornut#7735)

* Examples: undo adding SDL3 example to Visual Studio sln.

* Backends: OSX: build fix. Amend 32f9dfc

* ImGuiStorage: tweak impl for BuildSortByKey().

* Inputs: fixed using Shortcut() or SetNextItemShortcut() within a disabled block bypassing the disabled state. (ocornut#7726)

* Tables: moved TableGetHoveredColumn() to public API. (ocornut#7715, ocornut#3740)

* Windows: BeginChild(): fixed a glitch when during a resize of a child window which is tightly close to the boundaries of its parent. (ocornut#7706)

* Nav: store NavJustMovedToIsTabbing + shuffle a few nav related fields.

(for usage by multi-select)

* Backends: OpenGL2, OpenGL3: ImGui_ImplOpenGL3_NewFrame() recreates font texture if it has been destroyed by ImGui_ImplOpenGL3_DestroyFontsTexture(). (ocornut#7748)

Analogous to change to Vulkan backend in 1.90.

* Backends: Win32: Fixed warning with old MinGW/GCC versions.

* Drags: added ImGuisliderFlags_WrapAround flag for DragInt(), DragFloat() etc. (ocornut#7749)

* Fix typo, rename ImGuisliderFlags_WrapAround flag to ImGuiSliderFlags_WrapAround. (ocornut#7752, ocornut#7749)

* Checkbox: minor tidying up to simplify work on multi-select branch.

* Backends: Allegro5: Correctly handle unstable bit in version checks (ocornut#7755)

* Backends: SDLRenderer3: Update for SDL_RenderGeometryRaw() API changes.

* Backends: SDL3: update for SDL_SetTextInputRect() -> SDL_SetTextInputArea() api change. (ocornut#7760, ocornut#7754)

* Examples: SDL3: Remove use of SDL_HINT_IME_NATIVE_UI.

* Disabled: Reworked 1.90.8 behavior of Begin() not inheriting current BeginDisabled() state. Only tooltip are clearing that state. (ocornut#211, ocornut#7640)

* imgui_freetype: fixed divide by zero while handling FT_PIXEL_MODE_BGRA glyphs. (ocornut#7267, ocornut#3369)

* IO: do not claim io.WantCaptureMouse=true on the mouse release frame of a button which was pressed over void.  (ocornut#1392)

* Version 1.90.9

* Version 1.91.0 WIP

* Backends: SDL3: Update for API changes: SDLK_x renames and SDLK_KP_x removals (ocornut#7761, ocornut#7762)

Also updated function signature in SDL2 backend to match and because it is expected we will use that data (as per ocornut#7672)

* Backends: SDL3: Updated comments (IME seems fixed in SDL3). Added SDL3 examples to Visual Studio solution.

* Debug Tools: Added IMGUI_DEBUG_LOG(), ImGui::DebugLog() in public API. (ocornut#5855)

* Debug Log: Added "Configure Outputs.." button. (ocornut#5855)

* Backends: SDL3: add default case to fix warnings. (ocornut#7763)

* Demo: changed style editor inline block to its own window.

* IO: added io.PlatformOpenInShellFn handler to open a link/folder/file in OS shell, added IMGUI_DISABLE_DEFAULT_SHELL_FUNCTIONS. (ocornut#7660)

* (Breaking) IO, IME: renamed platform IME hook io.SetPlatformImeDataFn() -> io.PlatformSetImeDataFn() and added explicit context.

* Commented out obsolete ImGuiModFlags and ImGuiModFlags_XXX values (renamed to ImGuiKeyChord and ImGuiMod_XXX in 1.89). (ocornut#4921, ocornut#456)

* Build fix for non Windows platforms.

* IO: disable default io.PlatformOpenInShellFn() implementation on iPhone, as compiler errors that system() is not available on iOS.

* Internals: added FontScale storage.

* Added TextLink(), TextLinkOpenURL() hyperlink widgets. (ocornut#7660)

* Internals: added FontScale storage (amend 0f63d3e).

* Backends: GLFW,SDL2: Added ioPlatformOpenInShellFn handler for web/Emscripten versions. (ocornut#7660)

* IO: amend PlatformOpenInShellFn specs to return a bool. (ocornut#7660)

Amend 8f36798

* Misc tweaks, comments.

* TreeNode: rename/rework ImGuiNavTreeNodeData system to be usable by more features. (ocornut#2920, ocornut#1131, ocornut#7553)

Reworked to it is easier during TreeNode code to request extra data to be stored.

* Fixed Unix version of PlatformOpenInShellFn_DefaultImpl. (ocornut#7772, ocornut#7660)

+ Enable on non-iPhone macOS builds

* DemosFix typo in help text in demo Tables/Borders (ocornut#7780)

The help text for flags had a "V" flag duplicated, this change corrects it to the missing "H" flag.

* Backends: Win32: fixed ImGuiMod_Super being mapped to VK_APPS instead of VK_LWIN||VK_RWIN (ocornut#7768, ocornut#4858, ocornut#2622)

Amend 0755767

The `ImGui_ImplWin32_UpdateKeyModifiers()` function maps `ImGuiMod_Super` to `VK_APPS`, the "Application" key located between the Right Windows (Super) and Right Control keys on the keyboard, see https://conemu.github.io/en/AppsKey.html

This means that when using `ImGui::GetIO().KeySuper` to try to get the down state of the `VK_RWIN` or `VK_LWIN` keys, it'll always return FALSE when either of those keys are held down, and only return TRUE when `VK_APPS` is held down.

* Backends: GLFW+Emscripten: (Breaking) Renamed ImGui_ImplGlfw_InstallEmscriptenCanvasResizeCallback() to ImGui_ImplGlfw_InstallEmscriptenCallbacks(), added GLFWwindow* parameter. (ocornut#7647, ocornut#7600)

+ Fixed Emscripten warning when using mouse wheel on some setups.

* Backends: GLFW+Emscripten: Added support for GLFW3 contrib port. (ocornut#7647)

* Backends: GLFW+Emscripten: Fixed build (ocornut#7647)

* Examples: SDL3+OpenGL: Update for API changes: SDL_GL_DeleteContext() renamed to SDL_GL_DestroyContext().

* Fix definition check (ocornut#7793)

* Backends: SDL3: Update for API changes: SDL_GetProperty() change to SDL_GetPointerProperty(). (ocornut#7794)

* Backends: SDL3: fixed typo leading to PlatformHandleRaw not being set leading to SHOWNA path not working for multi-viewports.

* Internals: Added TreeNodeIsOpen() to facilitate discoverability. (ocornut#7553, ocornut#1131, ocornut#2958, ocornut#2079, ocornut#722)

* Added ImGuiDataType_Bool for convenience.

* Demo: Reworked "Property Editor" demo in a manner that more ressemble the tree data and struct description data that a real application would want to use.

* Added PushItemFlag(), PopItemFlag(), ImGuiItemFlags.

* (Breaking) Obsoleted PushButtonRepeat()/PopButtonRepeat() in favor of using new PushItemFlag()/PopItemFlag() with ImGuiItemFlags_ButtonRepeat.

* Added ImGuiItemFlags_AutoClosePopups as a replacement for internal's ImGuiItemFlags_SelectableDontClosePopup. (ocornut#1379, ocornut#1468, ocornut#2200, ocornut#4936, ocornut#5216, ocornut#7302, ocornut#7573)

* (Breaking) Renamed ImGuiSelectableFlags_DontClosePopups to ImGuiSelectableFlags_NoAutoClosePopups. (ocornut#1379, ocornut#1468, ocornut#2200, ocornut#4936, ocornut#5216, ocornut#7302, ocornut#7573)

* Obsoleted PushTabStop()/PopTabStop() in favor of using new PushItemFlag()/PopItemFlag() with ImGuiItemFlags_NoTabStop.

* Fixed pvs-studio warning.

* Demo: Property Editor: rearrange code + replace use of bool to proper ImGuiChildFlags.

Amend 46691d1

* Demo: Property Editor: add basic filter.

* Style: close button and collapse/window-menu button hover highlight made rectangular instead of round.

The reason they were round in the first place was to work better with rounded windows/frames.
However since the 4a81424 rework ocornut#6749 we can naturally use a tigher bounding box and it seems to work ok either way.

* Nav, Demo: comments.

* Clipper: added SeekCursorForItem() function, for use when using ImGuiListClipper::Begin(INT_MAX). (ocornut#1311)

Tagging ocornut#3609 just in case we made a mistake introducing a regression (but tests are passing and have been extended).

* TreeNode: Internals: facilitate dissociating item ID from storage ID (useful for 1861)

* Internals: rename recently added TreeNodeIsOpen() -> TreeNodeGetOpen(). (ocornut#7553, ocornut#1131, ocornut#2958, ocornut#2079, ocornut#722)

Amend ac7d6fb

* Backends: SDL3: Update for API changes: SDL_GetClipboardText() string ownership change. (ocornut#7801)

* MultiSelect: WIP range-select (ocornut#1861) (rebased six millions times)

* MultiSelect: Removed SelectableSpacing as I'm not sure it is of use for now (history insert)

* MultiSelect: Added IMGUI_HAS_MULTI_SELECT define. Fixed right-click toggling selection without clearing active id, could lead to MarkItemEdited() asserting. Fixed demo.

* MultiSelect: Demo sharing selection helper code. Fixed static analyzer warnings.

* MultiSelect: Renamed SetNextItemMultiSelectData() to SetNextItemSelectionUserData()

* MultiSelect: Transition to use FocusScope bits merged in master.

Preserve ability to shift+arrow into an item that is part of FocusScope but doesn't carry a selection without breaking selection.

* MultiSelect: Fix for TreeNode following merge of 011d475. Demo: basic test for tree nodes.

* MultiSelect: Fixed CTRL+A not testing focus scope id. Fixed CTRL+A not testing active id. Added demo code.

Comments.

* MultiSelect: Comments. Tweak demo.

* MultiSelect: Fix Selectable() ambiguous return value, clarify need to use IsItemToggledSelection().

* MultiSelect: Fix testing key mods from after the nav request (remove need to hold the mod longer)

* MultiSelect: Temporary fix/work-around for child/popup to not inherit MultiSelectEnabled flag, until we make mulit-select data stackable.

* MultiSelect: Fixed issue with Ctrl+click on TreeNode + amend demo to test drag and drop.

* MultiSelect: Demo: Add a simpler version.

* MultiSelect: Added ImGuiMultiSelectFlags_ClearOnEscape (unsure of best design), expose IsFocused for custom shortcuts.

* MultiSelect: Demo: Added pointer indirection and indent level.

This is to reduce noise for upcoming commits, ahead of adding a loop here.

* MultiSelect: Added ImGuiMultiSelectFlags_ClearOnClickWindowVoid. + Demo: showcase multiple selection scopes in same window.

* MultiSelect: Enter doesn't alter selection (unlike Space).

Fix for changes done in 5606.

* MultiSelect: Shallow tweaks/refactors.

Including moving IsFocused back internally for now.

* MultiSelect: Fixed needing to set RangeSrcPassedBy when not using clipper.

* MultiSelect: made SetNextItemSelectionData() optional to allow disjoint selection (e.g. with a CollapsingHeader between items). Amend demo.

* MultiSelect: Enter can alter selection if current item is not selected.

* MultiSelect: removed DragDropActive/preserve_existing_selection logic which seems unused + comments.

Can't find trace of early prototype for range-select but I couldn't find way to trigger this anymore. May be wrong. Will find out.

* MultiSelect: refactor before introducing persistant state pool and to facilitate adding recursion + debug log calls.

This is mostly the noisy/shallow stuff committed here, to get this out of the way.

* MultiSelect: (Breaking) Rename ImGuiMultiSelectData to ImGuiMultiSelectIO.

* MultiSelect: Demo tweak. Removed multi-scope from Advanced (too messy), made it a seperate mini-demo.

* MultiSelect: Internals rename of IO fields to avoid ambiguity with io/rw concepts + memset constructors, tweaks.

debug

* MultiSelect: (Breaking) Renamed 'RangeSrc -> 'RangeSrcItem', "RangeDst' -> 'RangeDstItem'

This is necessary to have consistent names in upcoming fields (NavIdItem etc.)

* MultiSelect: (Breaking) Renamed 'RangeValue' -> 'RangeSelected' + amend comments.

* MultiSelect: Remove ImGuiMultiSelectFlags_NoUnselect because I currently can't find use for this specific design.

And/or it seem partly broken.

* MultiSelect: Remove the need for using IsItemToggledSelection(). Update comments.

This is the simple version that past our tests. MultiSelectItemFooter() is in need of a cleanup.

* MultiSelect: Tidying up/simpllifying MultiSelectItemFooter().

Intended to be entirely a no-op, merely a transform of source code for simplification. But committing separatey from behavior change in previous change.

* MultiSelect: Clarify and better enforce lifetime of BeginMultiSelect() value.

* MultiSelect: Demo: first-draft of user-side deletion idioms.

(will need support from lib)

* MultiSelect: (Breaking) BeginMultiSelect() doesn't need two last params maintained by users. Moving some storage from user to core. Proper deletion demo.

* MultiSelect: Maintain NavIdSelected for user. Simplify deletion demo.

* MultiSelect: Further simplication of user code to support Deletion.

Provide standard RequestFocusItem storage.

* MultiSelect: Demo: Delete items from menu.

* MultiSelect: Fixed right-click handling in MultiSelectItemFooter() when not focused.

* MultiSelect: Cleanup unused comments/code.

* MultiSelect: (Breaking) Fix + Rename ImGuiMultiSelectFlags_NoMultiSelect to ImGuiMultiSelectFlags_SingleSelect as it seems easier to grasp.

Feature was broken by "Tidying up..." June 30 commit.

* MultiSelect: Comments, tweaks.

+ Alignment to reduce noise on next commit.

* MultiSelect: (Breaking) Use ImGuiSelectionUserData (= ImS64) instead of void* for selection user data.

Less confusing for most users, less casting.

* MultiSelect: move HasSelectionData to ImGuiItemFlags to facilitate copying around in standardized fieds.

Required/motivated to simplify support for ImGuiTreeNodeFlags_NavLeftJumpsBackHere (bc3c0ce) in this branch.

* MultiSelect: Tweak debug log to print decimal+hex values for item data.

Struggled to get standard PRIX64 to work on CI.

* MultiSelect: clear selection when leaving a scope with a nav directional request.

May need to clarify how to depends on actions being performed (e.g. click doesn't).
May become optional?

* MultiSelect: (Breaking) RequestSetRange's parameter are RangeFirstItem...RangeLastItem (which was always ordered unlike RangeSrcItem...RangeDstItme). Removed RangeDstItem. Removed RangeDirection.

* MultiSelect: Demo: rework ExampleSelection names to map better to typical user code + variety of Comments tweaks.

* MultiSelect: Demo: added simpler demo using Clipper. Clarify RangeSrcPassedBy doc.

* MultiSelect: (Breaking) Removed RangeSrcPassedBy in favor of favoring user to call IncludeByIndex(RangeSrcItem) which is easier/simpler to honor.

Especially as recent changes made it required to also update RangeSrcPassedBy after last clipper Step.
Should now be simpler.

* MultiSelect: Demo: rework ExampleSelection with an ExampleSelectionAdapter layer, allowing to share more code accross examples using different storage systems.

Not ideal way to showcase this demo but this is really more flexible.

* MultiSelect: Demo: Remove UserDataToIndex from ExampleSelectionAdapter.

Seems to make a better demo this way.

* MultiSelect: Demo: Make ExampleSelection use ImGuiID. More self-explanatory.

* MultiSelect: Demo: Deletion: Rework ApplyDeletionPreLoop to use adapter + fix PostLoop not using right value of RequestFocusItem.

Recovery made it transparent visually but user side selection would be empty for a frame before recovery.

* MultiSelect: Demo: Deletion: Various renames to clarify. Use adapter and item list in both ApplyDeletion functions.

This also minify the patch for an alternative/wip attmept at redesgining pre/post deletion logic. But turns out current attempt may be easier to grasp.

* Demo: Dual List Box: Added a dual list box (6648)

* MultiSelect: ImGuiMultiSelectIO's field are not used during loop anymore, stripping them out of comments.

* MultiSelect: moved RequestClear output so it'll match request list version better. Use Storage->RangeSrcItem in EndMultiSelect().

* MultiSelect: move shared logic to MultiSelectItemHeader().

No logic change AFAIK but added an indent level in MultiSelectItemHeader(). Logic changes will come in next commit.

* MultiSelect: Added ImGuiMultiSelectFlags_SelectOnClickRelease to allow dragging an unselected item without altering selection + update drag and drop demo.

* Demo: Assets Browser: Added assets browser demo.

* Demo: Assets Browser: store items, sorting, type overlay.

* MultiSelect: removed seemingly unnecessary block in BeginMultiSelect().

- EndIO.RangeSelected always set along with EndIO.RequestSetRange
- Trying to assert for the assignment making a difference when EndIO.RequestSetRange is already set couldn't find a case (tests passing).

* MultiSelect: clarified purpose and use of IsItemToggledSelection(). Added assert. Moved to multi-selection section of imgui.h.

* MultiSelect: added missing call on Shutdown(). Better reuse selection buffer.

* MultiSelect: (Breaking) io contains a ImVector<ImGuiSelectionRequest> list.

* MultiSelect: we don't need to ever write to EndIO.RangeSrcItem as this is not meant to be used.

* MultiSelect: added support for recovery in ErrorCheckEndWindowRecover().

* MultiSelect: use a single ImGuiMultiSelectIO buffer.

+ using local storage var in EndMultiSelect(), should be no-op.

* MultiSelect: simplify clearing ImGuiMultiSelectTempData.

* Demo: Assets Browser: add hit spacing, requierd for box-select patterns.

* MultiSelect: (breaking) renamed ImGuiMultiSelectFlags_ClearOnClickWindowVoid -> ImGuiMultiSelectFlags_ClearOnClickVoid. Added ImGuiMultiSelectFlags_ScopeWindow, ImGuiMultiSelectFlags_ScopeRect.

* MultiSelect: Box-Select: added support for ImGuiMultiSelectFlags_BoxSelect.

(v11)
FIXME: broken on clipping demo.

* MultiSelect: Box-Select: added scroll support.

* MultiSelect: Demo: rework and move selection adapter inside ExampleSelection.

* MultiSelect: added support for nested/stacked BeginMultiSelect().

Mimicking table logic, reusing amortized buffers.

* MultiSelect: remove ImGuiSelectionRequest/ImGuiMultiSelectIO details from public api to reduce confusion + comments.

* MultiSelect: move demo's ExampleSelection to main api as a convenient ImGuiSelectionBasicStorage for basic users.

* MultiSelect: reworked comments in imgui.h now that we have our own section.

* MultiSelect: Demo: Assets Browser: added deletion support. Store ID in selection. Moved QueueDeletion to local var to emphasis that this is a user extension.

* MultiSelect: Demo: Assets Browser: track scrolling target so we can roughly land on hovered item.

It's impossible to do this perfectly without some form of locking on item because as the hovered item X position changes it's easy to drift.

* MultiSelect: Box-Select: Fixed holes when using with clipper (in 1D list.)

Clipper accounts for Selectable() layout oddity as BoxSelect is sensitive to it.
Also tweaked scroll triggering region inward.
Rename ImGuiMultiSelectFlags_NoBoxSelectScroll to ImGuiMultiSelectFlags_BoxSelectNoScroll.
Fixed use with ImGuiMultiSelectFlags_SinglaSelect.

* MultiSelect: Box-Select: Added ImGuiMultiSelectFlags_BoxSelect2d support. Enabled in Asset Browser. Selectable() supports it.

* MultiSelect: Box-Select: Refactor into its own structure, designed for single-instance but closer to being reusable outside Multi-Select.

Kept same member names.

* MultiSelect: Box-Select: Refactor: Renames.

Split into two commits to facilite looking into previous one if needed.

* MultiSelect: Box-Select: Fixed scrolling on high framerates.

* MultiSelect: Box-Select: Further refactor to extra mode code away from multi-select function into box-select funcitons.

* MultiSelect: Fixed ImGuiSelectionBasicStorage::ApplyRequests() incorrectly maintaining selection size on SelectAll.

* MultiSelect: Comments + Assets Browser : Tweak colors.

* MultiSelect: Added ImGuiMultiSelectFlags_NoRangeSelect. Fixed ImGuiMultiSelectFlags_ScopeRect not querying proper window hover.

* MultiSelect: Box-Select: Fixed CTRL+drag from void clearing items.

* MultiSelect: Box-Select: Fixed initial drag from not claiming hovered id, preventing window behind to move for a frame.

* MultiSelect: Fixed ImGuiMultiSelectFlags_SelectOnClickRelease over tree node arrow.

* MultiSelect: (Breaking) merge ImGuiSelectionRequestType_Clear and ImGuiSelectionRequestType_SelectAll into ImGuiSelectionRequestType_SetAll., rename ImGuiSelectionRequest::RangeSelected to Selected.

The reasoning is that it makes it easier/faster to write an adhoc ImGuiMultiSelectIO handler (e.g. trying to apply multi-select to checkboxes)

* MultiSelect: Simplified ImGuiSelectionBasicStorage by using a single SetItemSelected() entry point.

* MultiSelect: Comments + tweaked location for widgets to test ImGuiItemFlags_IsMultiSelect to avoid misleading into thinking doing it before ItemAdd() is necessary.

* MultiSelect: Demo: make various child windows resizable, with synched heights for the dual list box demo.

* MultiSelect: added ImGuiMultiSelectFlags_NoAutoSelect, ImGuiMultiSelectFlags_NoAutoClear features + added Checkbox Demo

Refer to "widgets_multiselect_checkboxes" in imgui_test_suite.

* MultiSelect: Box-Select: fix preventing focus. amend determination of scope_hovered for decorated/non-child windows + avoid stealing NavId. (ocornut#7424)

* MultiSelect: Demo: use Shortcut().

Got rid of suggestion to move Delete signal processing to BeginMultiSelect(), seems unnecessary.

* RangeSelect/MultiSelect: (Breaking) Added current_selection_size to BeginMultiSelect().

Required for shortcut routing so we can e.g. have Escape be used to clear selection THEN to exit child window.

* MultiSelect: Box-Select: minor refactor, tidying up.

* MultiSelect: Box-Select: when dragging from void, first hit item sets NavId by simulating a press, so navigation can resume from that spot.

* MultiSelect: added GetMultiSelectState() + store LastSelectionSize as provided by user, convenient for quick debugging and testing.

* MultiSelect: Box-Select: fixed "when dragging from void" implementation messing with calling BeginMultiSelect() without a selection size.

* MultiSelect: (breaking) renamed ImGuiSelectionBasicStorage::AdapterData to UserData.

* MultiSelect: Box-Select: fixes for checkboxes support. Comments.

* MultiSelect: (breaking) renamed ImGuiMultiSelectFlags_BoxSelect -> ImGuiMultiSelectFlags_BoxSelect1d, ImGuiMultiSelectFlags_BoxSelect2d -> ImGuiMultiSelectFlags_BoxSelect.

ImGuiMultiSelectFlags_BoxSelect1d being an optimization it is the optional flag.

* MultiSelect: mark parent child window as navigable into, with highlight. Assume user will always submit interactive items.

* MultiSelect: (breaking) Added 'items_count' parameter to BeginMultiSelect(). Will enable extra features, and remove equivalent param from ImGuiSelectionBasicStorage::ApplyRequests(.

* MultiSelect: added ImGuiSelectionBasicStorage::GetStorageIdFromIndex() indirection to be easier on the reader.

Tempting to make it a virtual.

* MultiSelect: fixed ImGuiSelectionBasicStorage::Swap() helper.

* MultiSelect: added ImGuiSelectionExternalStorage helper. Simplify bool demo.

* MultiSelect: comments, header tweaks., simplication (some of it on wiki).

* MultiSelect: ImGuiSelectionBasicStorage: added GetNextSelectedItem() to abstract selection storage from user. Amend Assets Browser demo to handle drag and drop correctly.

* MultiSelect: ImGuiSelectionBasicStorage: rework to accept massive selections requests without flinching.

Batch modification + storage only keeps selected items.

* MultiSelect: ImGuiSelectionBasicStorage: simplify by removing compacting code (compacting may be opt-in?).

GetNextSelectedItem() wrapper gives us more flexibility to work on this kind of stuff now.

* MultiSelect: ImGuiSelectionBasicStorage: move function bodies to cpp file.

+ make ImGuiStorage::BuildSortByKey() less affected by msvc debug mode.

* Demo: Assets Browser: added a way to disable sorting and hide sorting options.

This is mostly designed to showcase that on very large sets (e.g. 1 million) most of the time is likely spent on sorting.

* MultiSelect: ImGuiSelectionBasicStorage: (breaking) rework GetNextSelectedItem() api to avoid ambiguity/failure when user uses a zero id.

* MultiSelect: provide RangeDirection to allow selection handler to handler backward shift+click.

* MultiSelect: ImGuiSelectionBasicStorage: added PreserveOrder, maintain implicit order data in storage.

Little tested but provided for completeness.

* MultiSelect: (breaking) renamed ImGuiMultiSelectFlags_BoxSelect -> ImGuiMultiSelectFlags_BoxSelect2d. Which include not assuming one flag imply the other.

Amend 2024/05/31 commit.

* MultiSelect: Shift+Tab doesn't enable Shift select on landing item.

* MultiSelect: added ImGuiMultiSelectFlags_NoAutoClearOnReselect + tweak flags comments. (ocornut#7424)

* MultiSelect: minor tidying up.

Checkbox() was reworked in master effectively fixing render clipping when culled by BoxSelect2d's UnclipMode.

* MultiSelect: added courtesy ImGuiMultiSelectFlags_NavWrapX flag so we can demo this until a nav api is designed.

* MultiSelect: Box-Select: uses SetActiveIdUsingAllKeyboardKeys() to avoid nav interference, much like most drag operations.

* MultiSelect: Box-Select: handle Esc to disable box-select.

This avoid remove a one-frame delay when finishing box-select, where Esc wouldn't be routed to selection but to child.

* MultiSelect: ImGuiSelectionBasicStorage: optimized for smaller insertion amounts in larger sets + fix caling batch select with same value.

* MultiSelect: Better document how TreeNode() is not trivially usable yet.

Will revert when the time is right.

* MultiSelect: added Changelog for the feature. Removed IMGUI_HAS_MULTI_SELECT.

* Demo: moved ExampleTreeNode, ExampleMemberInfo above in the demo file. Tidying up index.

+ change ExampleTreeNode::UID from ImGuiID to int to not suggest that the user ID needs to be of a certain type

* Demo: moved some fields inside a struct.

* Demo: moved menu bar code to its own function.

* MultiSelect: using ImGuiMultiSelectFlags_NoRangeSelect ensure never having to interpolate between two ImGuiSelectionUserData.

* Inputs: added SetItemKeyOwner(ImGuiKey key) in public API. (ocornut#456, ocornut#2637, ocornut#2620, ocornut#2891, ocornut#3370, ocornut#3724, ocornut#4828, ocornut#5108, ocornut#5242, ocornut#5641)

* Backends: SDL3: Update for API changes: SDL_GetGamepads() memory ownership change. (ocornut#7807)

* TabBar, Style: added style option for the size of the Tab-Bar Overline (ocornut#7804)

Amend 21bda2e.

* Added a comment hinting at how to set IMGUI_API for shared librairies on e.g. Linux, macOS (ocornut#7806)

* Demo: rework Property Editor.

* Nav: fixed c licking window decorations (e.g. resize borders) from losing focused item when within a child window using ImGuiChildFlags_NavFlattened.

In essence, using ImGuiFocusRequestFlags_RestoreFocusedChild here is a way to reduce changes caused by FocusWindow(), but it could be done more neatly.
See amended "nav_flattened" test.

* Demo: Property Editor: using ImGuiChildFlags_NavFlattened now that a bug is fixed. Fixed static analyzer.

* Backends: OpenGL3: Fixed unsupported option warning with apple clang (ocornut#7810)

* Internals, TreeNode: indent all render block into its own scope (aim is to add a is_visible test there later)

* Internals, TreeNode, Selectable: tweak span_all_columns paths for clarity.

* CollapsingHeader: left-side outer extend matches right-side one (moved left by one pixel)

Amend c3a348a

* Debug Log: fixed incorrect checkbox layout when partially clipped., doesn't parse 64-bits hex value as ImGuiID lookups.

* Groups, Tables: fixed EndGroup() failing to correctly capture current table occupied size. (ocornut#7543)

See "layout_group_endtable" test.

* MultiSelect: sequential SetRange merging not generally handled by box-select path, useful for others.

* MultiSelect: add internal MultiSelectAddSetAll() helper.

* MultiSelect: fixed an issue caused by previous commit.

Amend a285835. Breaks box-select.

---------

Co-authored-by: ocornut <omarcornut@gmail.com>
Co-authored-by: cfillion <cfillion@users.noreply.github.com>
Co-authored-by: Gary Geng <garygengxiao@gmail.com>
Co-authored-by: Martin Ejdestig <marejde@gmail.com>
Co-authored-by: Kevin Coghlan <mail@kevincoghlan.com>
Co-authored-by: Connor Clark <cjamcl@gmail.com>
Co-authored-by: Max Ortner <maxortner01@gmail.com>
Co-authored-by: Hugues Evrard <hevrard@users.noreply.github.com>
Co-authored-by: Aemony <10578344+Aemony@users.noreply.github.com>
Co-authored-by: Yan Pujante <ypujante@proton.me>
Co-authored-by: Cyao <94928179+cheyao@users.noreply.github.com>
Co-authored-by: wermi <32251376+wermipls@users.noreply.github.com>
Co-authored-by: Thomas Stehle <th.stehle@icloud.com>
@Maksons
Copy link

Maksons commented Aug 2, 2024

Hi, Nice work !

But " Build using --use-port=contrib.glfw3: demo"
with my system: Linux / Chrome (with enabled WebGPU)
I noticed problems:

  1. Do not resize to full-screen on "F11" press, it could be resized from Chrome menu -> full screen.
    resize back work on "F11" press.
    (old version "-sUSE_GLFW=3" work in both cases)

  2. Copy/Paste do not work in direction OtherApp (TextEditor/etc) -> Chrome
    work only in direction Chrome -> OtherApp
    (old version "-sUSE_GLFW=3" do not work copy/paste at all,
    but could work both directions with implemented extra functions I took here around)

@ypujante
Copy link
Contributor Author

ypujante commented Aug 2, 2024

Hi, Nice work !

But " Build using --use-port=contrib.glfw3: demo" with my system: Linux / Chrome (with enabled WebGPU) I noticed problems:

1. Do not resize to full-screen on "F11" press, it could be resized from Chrome menu -> full screen.
   resize back work on "F11" press.
   (old version "-sUSE_GLFW=3" work in both cases)

I never use F11 (I am on macOS) so I didn't know that was even a thing. I am not sure what triggers going fullscreen on F11 with -sUSE_GLFW=3. I would suggest having a menu and/or a button in your application to request full screen. If you look at my demo I have button(s) to go full screen. There is documentation about full screen support in the library.

I am curious what makes F11 work with the other implementation. The good news is that if this is an issue for you, you can still revert to using the old implementation as ImGui supports both (it is not a replacement).

2. Copy/Paste do not work in direction OtherApp (TextEditor/etc) -> Chrome
   work only in direction Chrome -> OtherApp
   (old version "-sUSE_GLFW=3" do not work copy/paste at all,
   but could work both directions with implemented extra functions I took here around)

Yes I am aware of the problem, I actually opened an issue for this purpose yesterday. Since the latest version of emscripten-glfw (which is not yet available in emscripten but can be still be used manually) I added an API to retrieve the "external" clipboard but I am not sure how to integrate it in ImGui, hence the ticket opened. You can actually see it in action in my live application: the code editor, which I control, uses this new API and as a result allow copy/paste from OtherApp -> Chrome...

But as you say, the old version does not support copy/paste at all, so it is already better :)

@ypujante
Copy link
Contributor Author

ypujante commented Aug 8, 2024

@Maksons I believe I understand why my library is "swallowing" F11. I will implement a way to allow enabling/disabling this kind of behavior in an upcoming release of emscripten-glfw

@ypujante
Copy link
Contributor Author

@Maksons I just wanted to let you know that I have released the new version of emscripten-glfw that addresses the 2 issues you mentioned a couple of weeks ago. I have submitted a PR in emscripten to merge the new version but unfortunately I do not control the timing of release.

If you want to use it before it gets released in emscripten, you can take a look at my other project, WebGPU Shader Toy, which has turned ImGui into an emscripten port and uses the latest version of emscripten-glfw, without depending on the emscripten release cadence...

Clipboard

If you are on PC or Linux, copy from Clipboard will just work as long as you use the keyboard shortcuts. On macOS you simply add this line (see Usage):

ImGui::GetIO().ConfigMacOSXBehaviors = emscripten::glfw3::IsRuntimePlatformApple();

Note that I am planning to open a PR on ImGui to do this automatically, but I need to wait for emscripten to a) merge my changes, b) release a version that contains the merged changes...

F11 key

See the Keyboard support section documentation for details, but in a nutshell, you simply add this to your code:

emscripten::glfw3::AddBrowserKeyCallback([](GLFWwindow* window, int key, int scancode, int action, int mods) {
  return mods == 0 && action == GLFW_PRESS && key == GLFW_KEY_F11;
});```

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Emscripten build runtime error: Unable to preventDefault inside passive event listener
3 participants