It doesn't make sense to ask everyone to file a bug if their
distribution is unsupported. We have manual build instructions for some
distributions and we can add instructions for more easily.
Android's OpenGL emulation layer (goldfish-opengl) has pre-processing
logic that looks for samplers of the type `samplerExternalOES` and
does a simple textual [replacement to change the type][1] to `sampler2D`
before compilation. It also [marks the sampler][2] as 'replaced' so
it can emulate the correct type at runtime.
However, this logic can lead to false positives when the sampler is
declared inside conditional macros. Hence, the sampler's type can be
incorrectly marked as `samplerExternalOES` even though the #if, #ifdef
conditional logic would have declared the type as `sampler2D`.
This seems to be a [known limitation][3].
WebRender (in particular the shared.glsl include) has such conditional
declaration of the texture units used from the shaders.
In particular, the sampler [sColor0 here][4] is declared within ifdefs
to have different types depending on the flags enabled, to allow the
shader to work with different image target kinds.
WebRender also maintain two versions of the compiled shaders in its cache:
1. An unoptimized version with all the conditional logic preserved
in the source until the shader is compiled at runtime on Android,
when the shader is actually used.
2. Multiple optimized versions for combinations of features required
These versions are produced during Servo [build][5]. Thus the
optimized versions eliminate most of the conditional declarations
at build time.
The bug in Servo with current code is because, [by default][6], WebRender
uses the *unoptimized* versions of the shaders. This means the conditional
GLSL source is evaluated at runtime by the Android emulator and thus ends
up with the incorrect type for the sColor0 sampler unit, which breaks the
[texture sampling in the fragment shader][7], causing it to always return
Vec4(0, 0, 0, 1) and rendering all elements on the page black.
This change forces WebRender to use the *optimized* version as a
workaround - the optimized versions have unconditional code for the
sampler declarations so are not susceptible to the emulator issue.
[1]: https://android.googlesource.com/device/generic/goldfish-opengl/+/refs/tags/android-platform-11.0.0_r40/system/GLESv2_enc/GL2Encoder.cpp#1644
[2]: https://android.googlesource.com/device/generic/goldfish-opengl/+/refs/tags/android-platform-11.0.0_r40/system/GLESv2_enc/GL2Encoder.cpp#1673
[3]: https://android.googlesource.com/device/generic/goldfish-opengl/+/refs/tags/android-platform-11.0.0_r40/system/GLESv2_enc/GL2Encoder.cpp#1571
[4]: b36399019c/webrender/res/shared.glsl (L206)
[5]: b36399019c/webrender/build.rs (L289)
[6]: b36399019c/webrender/src/renderer/init.rs (L214)
[7]: b36399019c/webrender/res/composite.glsl (L189)Fixes#31726.
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* clippy: remove extra static lifetimes in generated code
* clippy: fix return and cast in generated code
* clippy: fix more warnings in codegen
* clippy: fix all errors from generated files
* Update raw lags path for WPT import
I forgot to do this in #31616
* Update web-platform-tests to revision b'298d1599dbf6255aea63506daaa1702ff0c4fdc5'
---------
Co-authored-by: Samson <16504129+sagudev@users.noreply.github.com>
* Updated comment with 'layout_traits' to 'script_layout_interface'
* Rewrap text
* Remove trailing whitespace
---------
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
* Add support for partial and shallow clones for contributors with limited internet access
* Fix requested changes
* small typo
* Removed Trailing spaces and Fixed req changes
* Small fixes to the new documentation
---------
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
This adds basic support for `getClientRects()` by sharing code with the
implementation of `getBoundingClientRect()`. In addition to sharing
code, it also shares all of the bugs. Primarily, scrolilng positions are
not taken into account when return boundary rectangles.
The old logic was assuming that all whitespace was a break opportunity,
and that no newlines would be preserved.
Note that text shaping considers the advance of a newline to be the same
as a space. This was problematic because if we have a segment with a
preserved space and newline, only the advance of the space should
contrinute to the size of the block container. Therefore, I'm changing
the breaker logic in other to have newline characters in their own
segment.
Then glyph_run_is_whitespace_ending_with_preserved_newline can just be
renamed to glyph_run_is_preserved_newline.
This patch is still not perfect because it doesn't check allow_wrap(),
so `nowrap` is treated like `normal`, and `pre-wrap` like `pre`.
This brings the version of WebRender used in Servo up-to-date with Gecko
upstream. The big change here is that HiDPI is no longer handled via
WebRender. Instead this happens via a scale applied to the root layer in
the compositor. In addition to this change, various changes are made to
Servo to adapt to the new WebRender API.
Co-authored-by: Mukilan Thiyagarajan <mukilan@igalia.com>
These properties were being set to their initial value, this wasn't
needed because they are non-inherited properties and there isn't
anything else setting them.
Instead of using a simple `Atom` to identify a local font, use a data
structure. This allows us to carry more information necessary to
identify a local font (such as a path on MacOS). We need this for the
new version of WebRender, as fonts on MacOS now require a path.
This has a lot of benefits:
1. We can avoid loading fonts without paths on MacOS, which should
avoid a lot of problems with flakiness and ensure we always load the
same font for a given identifier.
2. This clarifies the difference between web fonts and local fonts,
though there is more work to do here.
3. This avoid a *lot* of font shenanigans, such as trying to work
backwards from the name of the font to the path of the font we
actually matched. In general, we can remove a lot of code trying to
accomplish these shenanigans.
4. Getting the font bytes always returns an `Arc` now avoiding an extra
full font copy in the case of Canvas.
