Issues occured (in dev build only) because a thread_local object from FMRSWRecursiveAccessDetector was used when destructing a delegate at exit time, but the thread_local object was already destructed.
Disabling the MT access detector for now for the problematic objects (the safest solution), but will investigate other solutions for 5.5.
#jira UE-210217, UE-210418
#tests Used with engine test and Lyra project. Ran Lyra in PIE, no regression noticed. Validated using breakpoint that delegates were called as expected.
#rb daren.cheng
[CL 32592225 by yohann dossantos in 5.4 branch]
Move icons into an editor folder so they don't get cooked
Change the override combo button to match the desired behavior
#jira UE-192063, UE-204831, UE-205036
#rb logan.buchy
[CL 32475269 by aditya ravichandran in 5.4 branch]
Hooked up overrides menu items which was added by [at]KJ.Jirak
#jira UE-204831
#rb aditya.ravichandran, logan.buchy
[CL 32475001 by maxime mercier in 5.4 branch]
It was for instance preventing the display of the fill of the progress bar, if shear/rotation was applyed on it and it was used in a retainer.
It's not using the change of the PR as it was working only in some case: the clipping zone we use always start from 0,0, so using the size of the Retainer (like in the PR) if it was smaller than the window would lead to the element being clipped out if the retainer was too far from 0,0.
#11455
#jira UE-206460
#tests In PIE, put a progress bar inside a retainer and played with shear and rotation on both the retainer and the progress bar: the blue fill is now always displayed. Engine test retainer test still succeed. Validated curve editor was still working as expected. Tested that animation are still played properly on windows (in case of drag n drop).
#rb daren.cheng
[CL 31521168 by batlam987 in 5.4 branch]
The intent of this change is twofold. First, it fixes a bug where the min/max/close buttons overlapped the row of tabs in windows with a main menu. Second, it reduces the vertical space consumed by our windows in the area around the title bar, main menu, and tabs.
* Reduced height of SVG images used for the min/max/restore/close window buttons from 34 pixels to 23 pixels. This avoids an overlap of the min/max/close buttons and the row of tabs in windows with a main menu.
* Rebalanced padding around the Unreal icon in windows with or without a main menu. There is now at most 1 pixel more spacing below the logo than above where before it could be 3 to 8 pixels more below than above.
* Reduced padding between tabs and main menu.
Note that these changes are all within #if !PLATFORM_MAC, so the Mac should not be affected.
#jira UE-204349
#rb brooke.hubert
[CL 31377272 by sebastian arleryd in 5.4 branch]
When Slate.EnableGlobalInvalidation=1, we go through FSlateElementBatcher::AddCachedElements and do BatchData->AddCachedBatches(CachedElementData.GetCachedBatches());
HDR viewports actually need to choose between BatchDataHDR and BatchData based on whether we want to composite ui with the scene later on
Added a flag on the batchdraw flags to specify which batchdata the viewport should end up into.
Since the flag might change when toggling, we invalidate the window the viewport is associated with whenever we detect a viewport change in FSlateRHIRenderer::ConditionalResizeViewport
Add GetViewportDynamicRange on ISlateViewport to be able to distinguish SDR and HDR viewports: we can end up in a scenario where we have an HDR window because still have an SDR viewport, like the FMovieViewport
The assumption is that we only play SDR movies, the code will have to be updated to support HDR movies
#jira UE-204214
#rb Patrick.Boutot
[CL 30964805 by benjamin rouveyrol in 5.4 branch]
FGraphAStar::PopIndex() isn't passing on its argument to the container - that existing behavior has been retained.
#rb james.hopkin
[FYI] Yoan.StAmant
[CL 30712410 by steve robb in ue5-main branch]
#tests: Tested in Editor and PIE, with and without artificial delay in generate tasks
#rnx <Release Notes exclude>
#rb yohann.dossantos
[CL 30682105 by viktor chlumsky in ue5-main branch]
With Ellipsis, all lines that are fully or partially visible will be displayed, the drawback of that being that if the "..." are displayed at the end of a line from which the lower part is clipped, the "..." are not visible.
With LineEllipsis, only lines that are fully visible vertically will be displayed. So if a line would be clipped at its lower part, the whole line will be culled, and the "..." will be displayed at the end of the last fully visible line, preventing the case where the "..." might not be seen.
#jira UE-195323
#tests Added new tests in the FontOverflow functionnal test from EngineTest. Visually checked in UnrealEditor that there were no regressions.
#rb Patrick.Boutot
#virtualized
[CL 30665613 by yohann dossantos in ue5-main branch]
To avoid aliasing on quad edges (for non SDF font only), the glyph quads are now drawn slighty bigger, with a larger UV area (by taking advantage of the padding, if there is any).
The rendered result will look exactly the same most of the time, but as soon as a scale or a non axis aligned rotation will be applyed to the render transform of the text, it will look a lot smoother on the edges.
By default, texts that are scaled up/down may now looks a bit different due to pixel rounding + different size of the quad.
Scrolling texts (without pixel snapping) move more smoothly.
Tweaked functionnal tests used for font to detect smaller local changes that got ignored otherwise, while they were important in the context of font. Maximum local error is now 0.5%.
#jira UE-188444
#tests Ran engine test font functionnal test and updated them. Compared the editor with and without the change. Compared scrolling text (pixel snapping disabled) with and without the change. Tolerance for local change is now
#rb Patrick.Boutot
#virtualized
[CL 30632576 by rcowsill in ue5-main branch]
Used the code from the PR as base, and changed it a little bit to optimize it and for readability.
Fix issue with leading tabs, but still need work to have it working properly for tabulation at any position in the text.
#jira UE-195689
#tests In editor, used test data with various amount of tabs at different position of the text. Multiple tabs at beggining of lines now works correctly, however, tabs at random position inside a text still don't aling the text properly (it won't be aligned to the next spacex4 multplier from the beggining of the line, but relatively to the last characters.
#rb Vincent.Gauthier
[CL 30632311 by emacser0 in ue5-main branch]
