change conversion to appunit from float of percentage calcuation #102
Conversation
Asun0204
changed the title
|
Since we want to test this before publishing a new release of app units, let's inline the code here for now. |
|
BTW, when creating a PR it's better not to use your main branch. |
Signed-off-by: Oriol Brufau <[email protected]>
Signed-off-by: Oriol Brufau <[email protected]>
There was a problem hiding this comment.
It may make more sense to modify this instead:
stylo/style/values/computed/length.rs
Lines 496 to 501 in 7e529e8
I thought that would be more likely to affect Firefox, but I have tried commenting that (and its callers) out, and Firefox compiles fine, so it's not used.
Asun0204
force-pushed
the
main
branch
2 times, most recently
from
6259dd8 to
8fd98ee
Compare
|
I'm sorry that the code you commit is overwitten by mistake, can you push the code again? @Loirooriol |
|
What code, 83dad25 ? |
|
Yes. And this opration that truncate in au construction not only in the percentage will expose another problem of #servo34714. The parameter of Timing Function style is f32 while it is f64 data type used in bezier function calculation. Loss can be produced in this progress. Could we just use truncate method in percentage, just like firefox? |
|
Sure, you can try that. But currently the logic first resolves LengthPercentage into a Length and then to an Au. You will need to remove the middle step. |
Signed-off-by: Asun0204 <[email protected]>
{"fail_fast": false, "matrix": [{"name": "Linux", "workflow": "linux", "wpt_layout": "none", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
Signed-off-by: Oriol Brufau <[email protected]>
{"fail_fast": false, "matrix": [{"name": "Linux", "workflow": "linux", "wpt_layout": "none", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
{"fail_fast": false, "matrix": [{"name": "Linux", "workflow": "linux", "wpt_layout": "none", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
Signed-off-by: Oriol Brufau <[email protected]>
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
Signed-off-by: Oriol Brufau <[email protected]>
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
I don't known how to remove the middle step. Do I need to copy the percenrage calculation in |
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
| @@ -562,8 +562,11 @@ impl LengthPercentage { | |||
| /// Convert the computed value into used value. | |||
| #[inline] | |||
| pub fn maybe_to_used_value(&self, container_len: Option<Au>) -> Option<Au> { | |||
| self.maybe_percentage_relative_to(container_len.map(Length::from)) | |||
| .map(Au::from) | |||
| if let Unpacked::Percentage(_) = self.unpack() { | |||
There was a problem hiding this comment.
Probably Unpacked::Calc should truncate too.
There was a problem hiding this comment.
Two wpt cases failed about calc, including calc-rounding-001.html and calc-rounding-003.html. Do we need to change Unpacked::Calc to truncate?
width: calc((100% - 4.5em) / 4); and width: calc((var(--width) - 4.5em) / 4); are compared in calc-rounding-001.html. But width: calc((var(--width) - 4.5em) / 4); is calcuated as a Length value instead of Calc.
There was a problem hiding this comment.
Test case of calc-rounding-001.html is also failed on firefox. The case on servo is passed but it looks differently than chrome and edge. So should we mark it failed as firefox or just modify percentage to truncate? @Loirooriol
Test case of calc-rounding-003.html has been already marked as expected: FAIL, I don't known why it's on the failed result.
There was a problem hiding this comment.
Rounding for calc() is better according to https://phabricator.services.mozilla.com/D202746
There was a problem hiding this comment.
The workflow that I just change the percenrage method using truncate has been passed. So should we merge app_units pr firstly? Then the stylo pr could be merged. @Loirooriol
https://github.com/Asun0204/servo/actions/runs/13385320860
Steps to merge this pr:
- The version of app_units is 0.7.7 now, we need upgrade to 0.7.8 and merge it with two interfaces.
- Publish app_units crate.
- We just merge this pr without upgrading version number because interfaces in stylo are not modified.
{"fail_fast": false, "matrix": [{"name": "Linux (WPT)", "workflow": "linux", "wpt_layout": "2020", "profile": "release", "unit_tests": false, "build_libservo": false, "bencher": false, "wpt_args": ""}]}
Asun0204
force-pushed
the
main
branch
4 times, most recently
from
3afe828 to
82d5920
Compare
…tion Signed-off-by: Asun0204 <[email protected]>
change conversion to appunit from float of percentage calcuation, prevent child box size is calculated too large.
Servo PR: servo/servo#34714