M117 public #197
Merged
M117 public #197
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
$ git describe canvaskit/0.38.2-1014-g26ce945468
m117: <include/core/SkSurfaceCharacterization.h> no longer available
commit 0bff57e0fe948bd9d96e92960e0ca993629c9a28
Author: Kevin Lubick <kjlubick@google.com>
Date: Fri Jun 9 14:29:17 2023 -0400
Move DDL and SkSurfaceCharacterization to include/private/chromium
This also renames them to be more clearly Ganesh-specific:
- SkDeferredDisplayList -> GrDeferredDisplayList
- SkDeferredDisplayListRecorder -> GrDeferredDisplayListRecorder
- SkSurfaceCharacterization -> GrSurfaceCharacterization
SkSurface::draw(sk_sp<GrDeferredDisplayList>...) has been moved
to a static function - skgpu::ganesh::DrawDDL. This was better than
re-writing the API to take a GrDeferredDisplayList* which would
require re-wrapping in a sk_sp to be passed into the deeper layers -
an error-prone and efficient strategy.
Shim headers are still around, as well as a shim over the old
SkSurface method until Chromium can be migrated.
This CL has no functional changes in how DDLs work, besides
removing the ability to specify an offset, which was not
implemented and there are no plans to do so.
Change-Id: Ibb263d6b90e3b3616ffd11c7a03b69e78dd02fe4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/706017
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Owners-Override: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Milestone 117
-------------
* `SkCanvas::flush()` has been removed. It can be replaced with:
```
if (auto dContext = GrAsDirectContext(canvas->recordingContext())) {
dContext->flushAndSubmit();
}
```
commit e3ca856e7ed6c285734dacf87faccdbf9a321f05
Author: Kevin Lubick <kjlubick@google.com>
Date: Wed Jul 26 08:58:05 2023 -0400
Remove legacy SkImage and SkSurface methods
Client CLs:
- http://ag/23168890
- http://ag/23171851
- http://ag/23473641
- http://ag/23494824
- http://cl/551209733
- https://crrev.com/c/4566404
- https://crrev.com/c/4705004
- flutter/engine#42425
- flutter/engine#43786
- flutter/engine#43965
Change-Id: I06c71cf6dc77aeaa3f78dec61c7b7c6eca688884
Bug: b/40045064
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/729776
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
commit f437d01005173a284d1e9c592bd0c61360a6cad1
Author: Kevin Lubick <kjlubick@google.com>
Date: Mon Jun 26 14:35:06 2023 -0400
Remove slug-related #ifdefs from src/core
This also enforces IWYU on some of the affected types
SkPicture etc.
I made SkPicturePlayback a friend of SkCanvas so the former
could call SkCanvas::draw(Slug) directly without needing to
have the implementation be linked in in a CPU build.
I added a null-returning implementation for deserialization
because there is still a connection between SlugImpl -> SubRunContainer
-> Ganesh code that means we cannot quite move slugs into
a CPU only build
Canary-Chromium-CL: 4649851
Bug: skia:14317
Change-Id: I98f776596288e2f6f3266549f5c9c8d12e9ba551
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/715416
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Mention SkCanvas::flush() is emulated
…er" on aarch64 build_Linux.sh has also been adjusted/re-arranged to insist on being re-run on aarch64.
|
Windows build is running out of space. I wonder what it requires... |
…ure on windows Trying to avoid windows running out of disk space.
…ing code just before removal Can't call other surface.flush() method either; use direct A lot of typos and missing namespaces typo
Milestone 114
-------------
* `SkImages::GetBackendTextureFromImage` has been renamed `SkImages::MakeBackendTextureFromImage`.
* `SkImage::getBackendTexture()` has been moved to `SkImages::GetBackendTextureFromImage()` in
`SkImageGanesh.h`.
* `SkImage::makeTextureImage()` has been moved to `SkImages::TextureFromImage()` in
`SkImageGanesh.h`.
* `SkImage::flush()` and `SkImage::flushAndSubmit()` has been moved to
`GrDirectContext::flush()` and `GrDirectContext::flushAndSubmit()` in `SkImageGanesh.h`.
…es::TextureFromImage`.
…bmit()`
Milestone 114
-------------
* `SkImage::flush()` and `SkImage::flushAndSubmit()` has been moved to
`GrDirectContext::flush()` and `GrDirectContext::flushAndSubmit()` in `SkImageGanesh.h`.
…Image()`
Milestone 114
-------------
* `SkImage::getBackendTexture()` has been moved to `SkImages::GetBackendTextureFromImage()` in
`SkImageGanesh.h`.
* `SkImage::makeTextureImage()` has been moved to `SkImages::TextureFromImage()` in
`SkImageGanesh.h`.
Also small white-space changes for uniformity.
…ictureRecorder::beginRecording() All the others have it. Without this, the example in issue 167 crashes on exit.
|
This skips about 10 fewer tests than 116b2 at this point , so is a marginal improvement. |
It appears that there are some subtle changes separating GPU-backed images vs
CPU-backed images, and GPU-backed images can't be compressed/saved directly.
Probably related to this:
m113 Bulk changes for SkImage->SkImages
commit 77472bf8434041b6e8a239f93816088e680f479c
Author: Kevin Lubick <kjlubick@google.com>
Date: Fri Mar 24 07:11:17 2023 -0400
Reland "Reland "Introduce SkImages namespace; remove Ganesh GPU code from SkImage_Raster""
This reverts commit 3e6bfdfea566b8368b10cd911499ae1bb09d3004.
Reason for revert: Fixing Android
…tureSize() instead.
Milestone 90
------------
* GrDirectContext::ComputeImageSize() is removed. Use SkImage::textureSize() instead.
https://review.skia.org/368621
https://review.skia.org/369317
https://review.skia.org/371958
static method and shared-pointers
|
Failing with "error: invalid command 'build_sphinx'". That seems to be a problem elsewhere? |
See: sphinx-doc/sphinx#11363 Remove setuptools support #11363
|
M118 is out. I 'll write a bit more in readme.m117 and bump it to 117b3 soon. Slightly undecisive about whether to merge readme.m117 or rename / prepend readme.m116, but probably less important as long as something is written down... anyway. |
…t/0.38.2-1015-gbeb4d7d77f
|
Bumped up to v117b3, and updated Readme (More breakage on m118. Sigh...) |
|
@kyamagu ready to go. |
|
New contributor @jonathanhogg , thanks. |
kyamagu
requested changes
Sep 22, 2023
kyamagu
approved these changes
Sep 22, 2023
|
@HinTak You can create a release on Github |
|
@kyamagu thanks - think it is done, going through CI. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
As the name says. M117 is out. The changes are relatively small and "obvious" (compared to m87 to m116...). The most worrying aspect is that surface.flushandsubmit and surface.flush are gone. I don't know if something else needs to be done. But I would recommend NOT merging this until that's investigated.
Canvas.flush is now emulated - but they actually tell you want to do in the m117 release note.
Also the aarch64 pre job is now split into two - one compiles 800 files for bundled 3rd party and another for skia proper, about 1200 files. Since worst case is about 5 hours 40 minutes, this split the worst case down to about 2 and 4. (Usually it is more like 1.5 hours and 2.5 hours, to make a total of 4).
Anyway, I am thinking of shifting that either sideway or behind the wheel matrix ie split the wheel matrix into native builds vs qemu builds, and re-order them into native builds, skia 3rd party, skia proper, and qemu builds, docs etc. It is more useful to see the result of native build earlier, for failures.
This should pass CI.