Make wlr_output_enable idempotent #3306

DemiMarie · 2021-10-27T08:11:42Z

Enabling or disabling an output should be idempotent. Previously, this
was not the case, causing bugs.

emersion · 2021-10-27T08:13:54Z

10:05:29 <emersion> output->pending.enabled only has a meaningful value if WLR_OUTPUT_STATE_ENABLED is set
10:05:36 <emersion> same for the rest of the output state
10:07:00 <emersion> that's the status quo -- i'm not against changing it, but there are gotchas
10:07:53 <emersion> e.g. if we keep the previous state in pending, there has to be an exception for the wlr_buffer field and the gamma tables -- we don't want to keep these locked/allocated after we're done with the commit

DemiMarie · 2021-10-27T09:08:22Z

10:05:29 <emersion> output->pending.enabled only has a meaningful value if WLR_OUTPUT_STATE_ENABLED is set
10:05:36 <emersion> same for the rest of the output state
10:07:00 <emersion> that's the status quo -- i'm not against changing it, but there are gotchas
10:07:53 <emersion> e.g. if we keep the previous state in pending, there has to be an exception for the wlr_buffer field and the gamma tables -- we don't want to keep these locked/allocated after we're done with the commit

The only time this change will have any effect is if wlr_output_enable is called twice on the same output, with different values of the parameter, before the output commits. Needless to say, this will basically never happen for typical rootful compositors. For rootless compositors such as the one I am working on, however, this happens a LOT. The result is that the output is left with the wrong value of output->pending.enabled, which causes problems later.

emersion · 2021-10-27T09:10:31Z

I don't understand which kind of issues you're running into. If WLR_OUTPUT_STATE_ENABLED is unset in the flags, you shouldn't even look at pending.enabled.

DemiMarie · 2021-10-27T09:11:24Z

I don't understand which kind of issues you're running into. If WLR_OUTPUT_STATE_ENABLED is unset in the flags, you shouldn't even look at pending.enabled.

I will do some testing locally.

DemiMarie · 2021-10-27T09:15:08Z

Without this commit my compositor trips over this assertion.

This is a glue file to allow integration with builds.sr.ht.

emersion · 2021-11-01T15:59:21Z

wlroots has migrated to gitlab.freedesktop.org. This pull request has been moved to:

https://gitlab.freedesktop.org/wlroots/wlroots/-/merge_requests/3306

For `required` to disable search the value needs to be of `feature` type. Checking `gles2` via `in` keyword returns a `bool` but `required: false` makes the dependency optional instead of disabled.

This allows compositors to handle touch pointer emulation manually, instead of having Xwayland do it [1]. [1]: https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/691

This struct contains additional information for session device change events, such as the DRM connector ID that has changed.

When a connector ID is specified in a hotplug event, don't scan all connectors. Only scan the connector that has changed.

Callers can access output->front_buffer instead.

The protocol uses a signed integer here, which is also what the wlr_input_method_v2_preedit_string struct provides to compositors from the input method protocol. Sway currently just passes those int32_t values directly to this function leading to an implicit conversion.

Closes #3324.

Removing an input device requires unlinking it from the list of all headless input devices. For that implement a destroy function.

We were send a protocol error if INTERLACED or BOTTOM_FIRST was set. This is incorrect for the zwp_linux_dmabuf_params.create code-path because this kills the client without allowing it to gracefully handle the error. We should only send a protocol error if the client provides a bit not listed in the protocol definition.

They are never used in practice, which makes all of our flag handling effectively dead code. Also, APIs such as KMS don't provide a good way to deal with the flags. Let's just fail the DMA-BUF import when clients provide flags.

A wlroots user can easily get confused and think that `cap` refers to wlroots buffer capabilities, not array capacity.

Co-authored-by: Simon Zeni <[email protected]>

Enabling or disabling an output should be idempotent. Previously, this was not the case, causing bugs.

ci: add .gitlab-ci.yml

d1b7567

This is a glue file to allow integration with builds.sr.ht.

emersion and others added 23 commits November 1, 2021 18:54

s/GitHub/GitLab/

8bc1086

render: completely disable gles2 if requested but libEGL is found

760e166

For `required` to disable search the value needs to be of `feature` type. Checking `gles2` via `in` keyword returns a `bool` but `required: false` makes the dependency optional instead of disabled.

xwayland: add support for -noTouchPointerEmulation

3e801d6

This allows compositors to handle touch pointer emulation manually, instead of having Xwayland do it [1]. [1]: https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/691

backend/session: introduce wlr_device_change_event

2ff4e11

This struct contains additional information for session device change events, such as the DRM connector ID that has changed.

backend/drm: handle per-connector hotplug events

e13f3f8

When a connector ID is specified in a hotplug event, don't scan all connectors. Only scan the connector that has changed.

output: fix leak of wlr_drm_format

8634dd3

CONTRIBUTING.md: add CoC section

fc1ed72

scene: fix calloc size mismatch

eb5f23d

text-input/input-method: handle strdup() failure

e326b76

export-dmabuf-v1: stop using wlr_output_export_dmabuf

f20c49d

screencopy-v1: stop using wlr_output_export_dmabuf

fa77aeb

output: drop wlr_output_export_dmabuf

76bab68

Callers can access output->front_buffer instead.

render/allocator: make wlr_allocator part of the public API

02a1ae1

egl: use alts for EGL_EXT_device_enum, if missing

3a685b1

util/token: don't leak /dev/urandom fd to children

4a8e681

Closes #3324.

backend/headless: unlink input device on destroy

8274c85

Removing an input device requires unlinking it from the list of all headless input devices. For that implement a destroy function.

Remove support for DMA-BUF flags

a04cfca

They are never used in practice, which makes all of our flag handling effectively dead code. Also, APIs such as KMS don't provide a good way to deal with the flags. Let's just fail the DMA-BUF import when clients provide flags.

Improve wlr_drm_format documentation

b5d4bc3

A wlroots user can easily get confused and think that `cap` refers to wlroots buffer capabilities, not array capacity.

output: add wlr_output_init_render

142d10e

Co-authored-by: Simon Zeni <[email protected]>

examples: init wlr_output with allocator and renderer

6d6e70b

tinywl: autocreate allocator and init output

a6538ce

bl4ckb0ne and others added 10 commits November 18, 2021 09:37

backend: remove backend ensure renderer and allocator check

0c76aef

backend: remove backend_get_allocator

6dc6af1

types/wlr_screencopy_v1: use renderer from output

d07c87f

types/wlr_scene: use renderer from wlr_output

5a98eae

backend/x11: get renderer from wlr_x11_output

5f11198

backend/headless: don't store the parent renderer

a143093

backend/drm: stop initializing backend renderer

42549a1

backend/multi: remove backend_get_renderer

d1ebd52

backend: remove wlr_backend_get_renderer

fdf3169

Make wlr_output_enable idempotent

7375cfa

Enabling or disabling an output should be idempotent. Previously, this was not the case, causing bugs.

DemiMarie force-pushed the idempotent-enable-disable branch from ab53f5e to 7375cfa Compare November 18, 2021 17:33

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make wlr_output_enable idempotent #3306

Make wlr_output_enable idempotent #3306

DemiMarie commented Oct 27, 2021

emersion commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

emersion commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

emersion commented Nov 1, 2021

Make wlr_output_enable idempotent #3306

Are you sure you want to change the base?

Make wlr_output_enable idempotent #3306

Conversation

DemiMarie commented Oct 27, 2021

emersion commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

emersion commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

DemiMarie commented Oct 27, 2021

emersion commented Nov 1, 2021