G
Thinking about the 0.5 offset - how is that handled in different implementations wrt to U8 buffers? Is 127 mapped to 0 or is 128 mapped to 0? Or is 127.5 mapped to 0? I guess that's even a relevant question for SVG renderers.
G
As for out-of-range values, Gemini claims that in 8- or 16-bit integer pipelines, values are clamped, while AE and blender 32-bit fp (ie HDR) pipelines leave values unclamped.
G
As for "How do people generate animated displacement maps? ", I got Procedural noise (turbulence) Animated shapes Physics simulation
G
What's that? Google says: Expand Output. A checkbox that grows the layer's bounding box automatically so displaced pixels are not clipped at the original layer borders.
G
The "0 shift midpoint" is throughout 0.5 or 128. The one exception is blender, where it seems to be adjustable between 0 and 1, with a default of 0.5 Here is a some affect effect tutorial: https://www.youtube.com/watch?v=J-goz3c5qvE
G
Another one: How do people create displacement maps? If you want to do something like LiquidGlass, you need to dynamically create a displacement map that matches the size of the thing you want to displace. How does one do that? I've only seen PNGs and JS code that do that, both of those approaches won't work - our CSS can't do JS and preloading breaks the moment you resize.
G
Another question: Did you/Gemini just not list the offset or do they all assume 0.5?
G
How do they handle colorspaces and out of [0..1] range values? option to expand output to accommodate all displaced pixels What's that? also: https://api.skia.org/classSkImageFilters.html#aa83bc85ebe76f2f6904a51907e63f876
G
feDisplacementMap: one map selectable X/Y coordinates: R, G, B, A single scale map is stretched to fit filter region default colorstate: linear RGB Inkscape displacement map: just the feDisplacementMap attributes Gimp displacement filter parameters: separate maps for X/Y selectable channels for X/Y: R, G, B, A polar mode in addition to cartesian, with various options: whirl/pinch, center edge modes interpolation: nearest/bilinear/bicubic/nohalo... After Effects (according to gemini): X/Y scale…
G
Add high-level apis for arithmetic, turbulence and displacement nodes. The render node constructors are still private.
G
I think we need to have a serious conversation
G
Add high-level apis for arithmetic, turbulence and displacement nodes. The render node constructors are still private.
G
We don't want that without proper API review.
G
Add high-level apis for arithmetic, turbulence and displacement nodes. The render node constructors are still private.
G
Add high-level apis for arithmetic, turbulence and displacement nodes. The render node constructors are still private.
G
Marge Bot (985fb0e1) at 16 Nov 22:22
G
Fallout from action-tree work and fixes a leak found along the way. Fixes: #8552 (closed)
G
Marge Bot (ddb840ab) at 16 Nov 22:22 Merge branch 'wip/chergert/n-actions-fix' into 'main' ... and 2 more commits
G
Summary Since 5a6802ca ("gtk: Use GtkActionTree throughout GTK"), a widget with no action node of its own still exports the AT-SPI Action interface, but its handlers do not answer: Properties.Get(org.a11y.atspi.Action, NActions): the getter returns NULL without setting an error, and GLib aborts the application. Any org.a11y.atspi.Action method: no reply is sent. The caller waits until its timeout. A plain GtkBox is enough. One atspi_action_get_n_actions() call from a screen reader kills the…
G
This branch couldn't be merged: Merging requests marked as auto-squash is not possible due to configuration. Please disable squashing or talk with the maintainers about the commit tagging config.
G
@chergert I wonder if we should revisit this, after all the recent textview work
G
A checklist of outstanding work here, from a recent irc discussion: Add a benchmark (see 'Blurred Images' in fishbowl) Do the same downscaling for shadows Implement "double tapping" Fix the irregularities in the screenshots above Figure out how to handle blurs when the input resolution does not match the output one Optimize repeated textures (which are commonly needed when blurring textures)
G
@Alex2772 Did you have a chance to get back to this?
G
I've taken a look at this and I'm not entirely clear on what factory we would want here to make things more flexible. The one thing I can come up with that might make sense is to allow setting a custom filter, instead of the hardcoded string filter. That would require some convention for how to feed the string from the search entry into the filter, something like /* Return the current string from the search entry, or NULL. Must only * be called from filter callbacks */ const char…
G
I think maybe @tyrylu needs to rebase his local branch on top of origin/main?
G
Hi @joaosacramento, how can I test this?
G
Closing this issue, since !10640 provides a fix for this issue.
G
Currently when starting up a gtk-demo or widget-factory on macos, it will show a window with decoration and transparent backend for a brief moment (around half a second) before the app has time to render its content for the first time. This MR addresses that by ensuring all code paths respect the existing show_on_next_swap flag, which already defers showAndMakeKey: until after the first buffer swap. Two paths were bypassing it: The key-window selection idle callback would call…
G
Closing this issue, since !10667 (merged) has been merged instead.