You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@abarth says this is the same root bug as #5203 and is a bug in Skia that the returned bounds rect for shadows is wrong. We need to make a reduced test case and file.
@chinmaygarde can you take on the task of making a reduction for Skia? I want to make sure that we get this on Skia's radar ASAP as I've had at least one person point this out to me when demoing flutter gallery to them.
Sorry it took so long for this one. Went on a wild goose chase looking in Skia and the framework. Its the raster cache not accounting for the shadow. Skia is fine.
We aren't suggesting that the cullRect of the SkPicture should include any extensions introduced to shapes as a result of the filters applied on the paints that were used at the time of the draw of said shapes on the canvas are we? Seems like a pretty invasive change.
Well, so far, there is no raster cache fix. But the issue is in there. #5203 is also related to the raster cache but its a different issue where the results of to a texture subsequently rendering that texture to the surface vs rendering directly to the window surface (or parent picture) is resulting in rendering artifacts.
We aren't suggesting that the cullRect of the SkPicture should include any extensions introduced to shapes as a result of the filters applied on the paints that were used at the time of the draw of said shapes on the canvas are we? Seems like a pretty invasive change.
There's code in Skia that knows how much filters outset the bounds of the geometry they're filtering. I thought that was wired up to the cull rect, but maybe it comes out somewhere else? If the cull rect doesn't take the filter outsets into account, then it can incorrectly cull pictures that actually would intersect a clip rect due to filter outsets.
This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of flutter doctor -v and a minimal reproduction of the issue.
Activity
eseidelGoogle commentedon Aug 5, 2016
@abarth says this is the same root bug as #5203 and is a bug in Skia that the returned bounds rect for shadows is wrong. We need to make a reduced test case and file.
eseidelGoogle commentedon Aug 11, 2016
@chinmaygarde can you take on the task of making a reduction for Skia? I want to make sure that we get this on Skia's radar ASAP as I've had at least one person point this out to me when demoing flutter gallery to them.
chinmaygarde commentedon Aug 22, 2016
Sorry it took so long for this one. Went on a wild goose chase looking in Skia and the framework. Its the raster cache not accounting for the shadow. Skia is fine.
chinmaygarde commentedon Aug 22, 2016
We aren't suggesting that the
cullRectof theSkPictureshould include any extensions introduced to shapes as a result of the filters applied on the paints that were used at the time of the draw of said shapes on the canvas are we? Seems like a pretty invasive change.eseidelGoogle commentedon Aug 22, 2016
@chinmaygarde I don't know. That's an @abarth question. Does your raster cache fix resolve #5203 as well?
chinmaygarde commentedon Aug 22, 2016
Well, so far, there is no raster cache fix. But the issue is in there. #5203 is also related to the raster cache but its a different issue where the results of to a texture subsequently rendering that texture to the surface vs rendering directly to the window surface (or parent picture) is resulting in rendering artifacts.
sethladd commentedon Aug 26, 2016
FWIW the shadow flash happens when going into or out of a component demo, too.
abarth commentedon Aug 29, 2016
There's code in Skia that knows how much filters outset the bounds of the geometry they're filtering. I thought that was wired up to the cull rect, but maybe it comes out somewhere else? If the cull rect doesn't take the filter outsets into account, then it can incorrectly cull pictures that actually would intersect a clip rect due to filter outsets.
39 remaining items
eseidelGoogle commentedon Apr 25, 2017
Progress continues. Almost done. Blocked on more fixes from skia.
eseidelGoogle commentedon May 2, 2017
Any status updates @chinmaygarde?
chinmaygarde commentedon May 2, 2017
Paths and text blobs still don't report their cull rects correctly. Waiting for fixes for the same. Will update this ticket when those changes land.
eseidelGoogle commentedon May 9, 2017
@chinmaygarde I presume this should get bumped from M2 and will end up as a May bug since it's unlikely to land today before we cut
alpha?abarth commentedon Jun 7, 2017
/cc @pylaligand
abarth commentedon Jun 7, 2017
@eseidelGoogle @chinmaygarde I thought this bug was fixed? @pylaligand seems to have just run into it again.
chinmaygarde commentedon Aug 28, 2017
This issue has been fixed. I should have closed this ticket a while ago.
github-actions commentedon Sep 4, 2021
This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of
flutter doctor -vand a minimal reproduction of the issue.