mirror of
https://github.com/bryanthaboi/gen1recomp.git
synced 2026-08-14 10:13:33 +02:00
bryanthaboi
803f86d5b3
* pop up a fake save * add CI to dev branch * Make ROM-free test tiers actually run on Windows (#267) Suite discovery, the extension-point catalog scan, mod test-dir pickup, and the meta-coverage corpus all shelled out to ls/find/test -d, which do not exist in cmd.exe. Every listing came back empty on Windows, so tiers ran 0 suites and still reported ALL TESTS PASSED. Add portable probes to tests/fs_io.lua (same Unix commands on Linux/macOS; dir /b and a shell-free rename-self existence check on Windows) and rewire the four call sites to them. 15/15 engine suites and 2/2 modkit suites now genuinely run and pass on Windows. Fixes #266 Co-authored-by: johnjohto <johnjohto@users.noreply.github.com> * intro api upgrade (#294) * intro api upgrade * api updates * fix tests * updated templates * android fixes for mods and saves (#297) * Android SCALING fixes (#298) * android fixes for mods and saves * perhaps this is the true scaling android issue fix --------- Co-authored-by: johnjohto <johtoboy@atomicmail.io> Co-authored-by: johnjohto <johnjohto@users.noreply.github.com>
46 lines
2.3 KiB
Lua
46 lines
2.3 KiB
Lua
-- Render targets measured in real framebuffer pixels.
|
|
--
|
|
-- love.graphics.newCanvas defaults its `dpiscale` to
|
|
-- love.graphics.getDPIScale(), so on a highdpi surface the canvas texture is
|
|
-- NOT the size it was asked for: newCanvas(160, 144) on a device reporting a
|
|
-- DPI scale of 2.755 allocates a 441x397 texture, and the 160x144 GB scene
|
|
-- then renders into it at 2.755 texels per GB pixel. conf.lua sets
|
|
-- t.window.highdpi on Android/iOS (required for Retina), and Android's
|
|
-- DisplayMetrics.density is routinely non-integer (1.5, 2.75, ...), so this
|
|
-- is the normal mobile case, not an edge case.
|
|
--
|
|
-- That fractional render breaks the whole premise of Renderer's integer
|
|
-- scale pipeline. Renderer:fitScale() picks a whole number of physical
|
|
-- pixels per GB pixel (7 on 1080p) and the composite blit lands on it
|
|
-- exactly -- but if the *source* already holds 2.755 texels per GB pixel,
|
|
-- some GB pixels are 2 texels wide and some are 3, and the nearest-neighbour
|
|
-- upscale turns them into 5 / 7 / 8 physical pixels instead of a uniform 7.
|
|
-- Measured on a 1080p density-2.755 fixture, only 43 of 160 columns came out
|
|
-- the right width. Fonts show it worst: their strokes are single pixels.
|
|
-- That is issue #208 ("some are noticeably stretched, especially fonts").
|
|
--
|
|
-- Forcing dpiscale = 1 makes canvas texels the pixels the renderer already
|
|
-- believes it is drawing. getWidth()/getHeight() are unaffected (they always
|
|
-- reported the requested size), so no geometry anywhere changes -- only the
|
|
-- resolution of the target. Desktop is unchanged; dpiscale is already 1
|
|
-- there because conf.lua sets highdpi on mobile only.
|
|
--
|
|
-- Exception: a canvas that is deliberately sized in LOVE *units* and blitted
|
|
-- back at unit scale 1 (Renderer's presentCanvas) must keep the screen's DPI
|
|
-- scale so its texture still covers the framebuffer; it is not built with
|
|
-- this helper.
|
|
|
|
local PixelCanvas = {}
|
|
|
|
-- One framebuffer pixel per w/h unit, always. `filter` is applied only when
|
|
-- given, so callers that relied on LOVE's default ("linear") keep it.
|
|
function PixelCanvas.new(w, h, filter)
|
|
local canvas = love.graphics.newCanvas(w, h, { dpiscale = 1 })
|
|
if filter and canvas and canvas.setFilter then
|
|
canvas:setFilter(filter, filter)
|
|
end
|
|
return canvas
|
|
end
|
|
|
|
return PixelCanvas
|