fix(layer-shell-scaling): frozen buffers must scale to output scaling. #112
Conversation
|
and now in the code , the way to use slurp is a wrong way |
Decodetalkers
force-pushed
the
fixclippy
branch
from
24bcd36
to
2e951ab
Compare
|
oh.. clippy is merged.. emm, so it is the problem that I have not find my pr |
Decodetalkers
force-pushed
the
fixclippy
branch
from
2e951ab
to
3ee9b64
Compare
|
Scratch that, I think we already have pixel perfect version from the comp in the buffer. We just need to set the scale hint. |
the size which is used to scale the buffer only accept int, so when the scale is |
Shinyzenith
changed the title
I sadly don't. As far as I'm aware, screencopy already returns to us a pixel-perfect version which is scaled appropriately due to Andreas's patches, we just need a way to hint to the compositor that we have the scaled version already. I'm not sure if my understanding is correct. I'll take a look at the code in the evening and maybe ask in sway-irc if needed. |
Decodetalkers
force-pushed
the
fixclippy
branch
from
3ee9b64
to
d1c1804
Compare
There was a problem hiding this comment.
Like @Shinyzenith says (and see my other review comment), the approach is incorrect. We already have a pixel perfect buffer that needs no changes.
Tried this on my machine and it only results in an even blurrier version of what happened before which was already incorrect.
libwayshot/src/lib.rs
Outdated
| let buffer = { | ||
| let output_size = output_info.logical_region.inner.size; | ||
| if output_size != frame_copy.frame_format.size { | ||
| let file = tempfile::tempfile().unwrap(); |
There was a problem hiding this comment.
| let file = tempfile::tempfile().unwrap(); | |
| let file = tempfile::tempfile()?; |
If not the correct error type, wrap it in another type, don't unwrap.
libwayshot/src/lib.rs
Outdated
| let image = image::imageops::resize( | ||
| &image, | ||
| output_size.width, | ||
| output_size.height, | ||
| image::imageops::FilterType::Triangle, | ||
| ); | ||
| init_overlay(&image, &file); | ||
| let pool = shm.create_pool( | ||
| file.as_fd(), | ||
| frame_copy | ||
| .frame_format | ||
| .byte_size() | ||
| .try_into() | ||
| .map_err(|_| Error::BufferTooSmall)?, | ||
| &qh, | ||
| (), | ||
| ); | ||
|
|
||
| pool.create_buffer( | ||
| 0, | ||
| output_size.width as i32, | ||
| output_size.height as i32, | ||
| (output_size.width * 4) as i32, | ||
| frame_copy.frame_format.format, | ||
| &qh, | ||
| (), | ||
| ) |
There was a problem hiding this comment.
There should be no reason to resize. We have a pixel perfect buffer already that represents every single pixel shown on the screen.
Why do you think you need to do this?
There was a problem hiding this comment.
when the scale is int, I will not resize it, but when it is fractal, I will resize it manully. Now on your screen it will not be blur anymore. The image will only be blur when the scale is set as 1.2, then I will resize it and make it a little blur, then it can be used to select area. this place it is just used to show the image and let user to select area
| let image_buffer = if let Some(slurp_region) = cli.slurp { | ||
| let slurp_region = slurp_region.clone(); | ||
| let image_buffer = if cli.slurp { | ||
| wayshot_conn.screenshot_freeze( | ||
| Box::new(move || { | ||
| || -> Result<LogicalRegion> { | ||
| let slurp_output = Command::new("slurp") | ||
| .args(slurp_region.split(' ')) | ||
| .output()? | ||
| .stdout; |
There was a problem hiding this comment.
Why change this behaviour? This is removing the ability to pass arguments to slurp.
There was a problem hiding this comment.
slurp only accept '-d' to select screen, but -d cannot be passed into, and it is useless
There was a problem hiding this comment.
all the argument is like -d -w, how can you pass the argument into it? I have try send it , but it can never be received
like such kind of error
There was a problem hiding this comment.
cargo run -- --slurp '\-d'
this command will work
but '-d' will never work
and wrong use of slurp, slurp only accept args like `-d`, it do not accept region, just run -d is enough
I have read the portocol of fractional, the protocal just accept a scale value to client and let client to adjust the scale. so finally the way to scale is done by client. So the scale should be done by ourself