Age | Commit message (Collapse) | Author | Files | Lines |
|
Fixes regression in 80f76c3aae438f7b9a7c1359c3f37aac460f934b. The
scissor is no longer always enabled, but since it was enabled on startup
there was visible clipping when resizing the window right after startup.
|
|
Only call `SDL_ShowWindow` once, since it seems old SDL versions (like
2.30.0, shipping with Ubuntu 24.04) do not allow a maximized window to
be minimized when `SDL_ShowWindow` is called each frame.
|
|
By using SDL logging, we can rely on SDL to filter the log messages by
their priority before they get sent to our custom output function. For
example, we now no longer get entries for created and destroyed Window
instances by default, since these have only "debug" priority whereas the
default priority for the "app" category is "info".
We can also get log messages raised by SDL itself in our log now. Log
levels can be controlled per category through the "SDL_LOGGING"
environment variable.
Many existing log messages have been assigned a category based on their
existing prefix.
|
|
Otherwise, at least on X11, the window is (visibly) re-created once the
SDL renderer is created. Also the window icon can be seen changing on
startup. Showing the window only after all related setup has been
completed avoids this.
On X11 this appears to have the added benefit of putting a "windowed
fullscreen" window on the active screen (in terms of where the mouse is)
rather than always using the primary screen.
On Windows, showing the window only after the initial paint avoids the
window initially appearing with its contents entirely black.
The window is still immediately shown in the following cases:
* In exclusive fullscreen mode, because on Windows this caused the
initial show of a fullscreen window to take much longer and on X11 it
could cause the resolution of the primary screen to be changed while
maximizing the Mana window on the active screen.
* In "windowed fullscreen" mode when the legacy OpenGL renderer is used,
where it was causing an additional flicker on Windows.
|
|
We don't need to keep an array of SubImage instances, but can just
remember the top, left, right and bottom margins and use those when
rendering the scaled ImageRect.
Graphics::drawRescaledImagePattern had to be extended to allow
specifying the source rectangle.
|
|
Clipping has been disabled globally by taking it out of
Graphics::pushClipArea. Now its name is a little confusing, but it can't
just be changed since it is part of Guichan.
Widgets that do need to clip their children use the new
Graphics::pushClipRect, which pushes a clipping rectangle without
affecting the local coordinates. These are:
* ScrollArea
* TextField
* TabbedArea (scrolling tabs)
* MiniMap
While it might count as a small optimization, I'm actually disabling
clipping because it is not always desired. For example it gets in the
way of rendering the complete ResizeGrip in the Jewelry theme because
that's a child widget.
|
|
|
|
Helps making sure we're not modifying the images in the rendering code.
Re-applies 4eea727b7649726670d8963d11ab4fd429624b3e (and reverts
363f71157a8107190b3bd2ba656faf0a0e63ab36).
|
|
* TrueTypeFont class now takes into account the graphics scale, in order
to render an appropriate higher-resolution texture.
* Removed TrueTypeFont::fontCounter, since TTF_Init/TTF_Quit already
keep a counter.
* Avoid copying the rendered string needlessly, when it already exists
in the cache. Avoid another copy, when inserting a new chunk into the
cache.
|
|
This can be used for smoother mouse cursor movement when rendering our
own mouse cursor (already changed in this commit) and is also necessary
for implementing support for HiDPI font rendering.
Also dropped some almost duplicated OpenGL code.
|
|
* Added "Scale" user option, which can either by "Auto" or an explicit
scaling factor. Its maximum value depends on the current resolution.
The "Auto" factor is based on keeping the logical resolution on at
least 800x600, wheres the maximum scale is based on keeping the
logical resolution on at least 640x480.
* Enabled support for High DPI. This means the rendering target can now
have a different resolution than the window size, which can happen on
macOS, Windows and Wayland. The resulting scale is multiplied by the
above user-controlled scale.
Currently, this looks ugly for non-integer scales, which are not used
on macOS and can only be configured on some Wayland compositors.
Has not been tested on Windows.
* Simplified OpenGL initialization (moved out of _beginDraw).
* Made sure _beginDraw/_endDraw sets a clip area also for SDLGraphics.
|
|
* Don't needlessly store or return raw pointers in BeingInfo
* Less copying, more moving
* Less else after return
* Make AddDEF a template instead of a macro
* Removed some unused includes
* Use range-based for loops
|
|
The configuration and setup UI were adjusted to the new options.
This also fixes issues in applying new video settings.
Default resolution was changed from 800x600 to 1280x720. VSync is
enabled by default while FPS limit was disabled.
Display aspect ratio for the resolution options.
I had to work around some macOS issues:
* Don't change window size when it appears to be "maximized", since it
just changes the rendering area while leaving the window maximized.
* Unset fullscreen display mode temporarily to allow changing resolutions,
otherwise the rendering area no longer matches the screen and mouse
input is also off.
* Removed SDL_WINDOW_ALLOW_HIGHDPI for now because it causes issues on
macOS, since we're not actually handling the scaling factor.
A Video class and an SDLGraphics subclass were split off from Graphics.
This setup has Less duplication and leaves the OpenGLGraphics and
SDLGraphics better separated.
Fixes #57
Fixes #58
|
|
modernize-use-auto
modernize-use-nullptr
modernize-use-override
modernize-use-using
|
|
|
|
Actually neither glFlush nor glFinish need to be called manually [1].
However, the rendering pipeline is then free to queue up future frames
which can cause a lag which is especially noticable with mouse movement.
To avoid this lag, we call glFinish after SDL_GL_SwapBuffers. This makes
sure processing of the next frame does not start before the current
frame is being displayed.
[1] http://www.opengl.org/wiki/Common_Mistakes#glFinish_and_glFlush
Reviewed-by: Yohann Ferreira
Sounded-fine-to: Erik Schilling
|
|
If the graphics driver supports this, there is no need to create
textures with power-of-two dimensions. It is then also preferred to use
regular textures than relying on the older GL_ARB_texture_rectangle
extension.
Reviewed-by: Erik Schilling
|
|
* Removed empty and unused function `setTargetPlane`
* Removed duplicated enabling of `GL_VERTEX_ARRAY`
* Removed strange smoothing code from `drawRescaledImage`
Reviewed-by: Erik Schilling
|
|
Textures were not rendering correctly on cards that did not support the
GL_ARB_texture_rectangle extension, since the alternative code path was
passing ints as floats and floats as ints to OpenGL.
Reviewed-by: Erik Schilling
|
|
|
|
Window positions are semi-smartly corrected as a result of the resize.
Not supported when using OpenGL on Windows for now.
Reviewed-by: Yohann Ferreira
|
|
Reviewed-by: Thorbjørn Lindeijer
Reviewed-by: Yohann Ferreira
|
|
This fixes overlay effects that are meant to scale with screen
resolution.
The problem was that the texture coordinates were not calculated
correctly. They were adjusted to the scaled size of the image, and when
scaling both the vertex and the texture coordinates, the image will
simply not appear scaled at all. Now the texture coordinates are
calculated based on the visible part of the original texture.
There was also a problem with the vertex coordinates, which were not
taking into account the visible part of the image.
TMW-Mantis-issue: 1047
|
|
Resolves: Mana-Mantis #265.
Trivial fix.
|
|
Use batch commands and not selecting already selected texture.
Signed-off-by: Jared Adams <jaxad0127@gmail.com>
|
|
|
|
|
|
Signed-off-by: Tametomo <irarice@gmail.com>
|
|
Also added the update copyright tool from the Wormux Team.
( And not forgetting credit's due. :P )
|
|
OpenGLGraphics::setVideoMode to stop Graphics::getWidth and getHeight returning 0
|
|
Also remove some extra new lines and fix eAthena's PartyTab define guards.
|
|
|
|
|
|
|
|
|
|
The current login wallpaper (as any streched ones) will look a bit less ugly...
|
|
resolution isn't the default.
You'll have to add this in map properties, for instance if you're want to keep ratio on overlay 0:
<map version="1.0" orientation="orthogonal" width="128" height="128" tilewidth="32" tileheight="32">
<properties>
...
<property name="overlay0keepratio" value="true"/>
...
</properties>
</map>
|
|
The SDL methods to rescale the wallpaper has been optimized to permit rescaling
at load time while OpenGL draws directly rescaled. Does someone know how to smooth
the rescaled image under OpenGL?
|
|
Shouldn't affect performance.
|
|
Conflicts:
Many files.
|
|
Reviewed-by: trapdoor
|
|
the open source project Wormux. To improve SDL performance, the number
of layers that are pushed out to the hardware or software buffers should
be reduced, which is where this function comes into play, as it combines
two surfaces together so that the number of blit operations is reduced.
This function is currently not used, but will be used once a good way to
link each of the target systems is determined so that it only initiates
when SDL is enabled, as well as making sure that each hook that uses
this function is benefiting from it sufficiently. At the moment, it's
suspected that the particle engine will likely be the most likely to
benefit from this function, followed by tile drawing, then sprite drawing.
Signed-off-by: Ira Rice <irarice@gmail.com>
|
|
d654758ef63f6515d678ceaf77d63a2693e08fb7, but for SDL instead. This
currently doesn't buy too much, but it's a little better than it used to
be.
TODO: Find out why SDL is bottlenecked, and try to bring its performance
up to OpenGL levels, if possible.
Signed-off-by: Ira Rice <irarice@gmail.com>
|
|
|
|
that position. This makes it so that when resolutions are changed, the
default locations stay relative to the window's position, and not the
800x600 screen resolution.
Signed-off-by: Ira Rice <irarice@gmail.com>
|
|
useful since buddy lists are tracked through the player relation
interface instead)
Signed-off-by: Ira Rice <irarice@gmail.com>
|
|
Signed-off-by: Ira Rice <irarice@gmail.com>
|
|
This dates back to the old days of TMW, but the usage instructions of
GPLv2 don't mention this being necessary. Since it doesn't add anything,
avoid the branding in these sections.
|
|
(cherry picked from mainline commit f6f8b2f885145d4cacf7a6c42d88b6a39f71b366)
Conflicts:
src/openglgraphics.cpp
|
|
|