Add debug option to display widget ids #876

cmyr · 2020-04-24T17:33:19Z

This will paint the id of each widget of some subtree in the bottom
right of that widget's bounds; useful for debugging various bits
of event flow and widget behaviour.

luleyleo · 2020-04-24T18:37:41Z

Good idea 👍

But this can cause overlapping ids:

Not sure how to deal with this but I would not like it to stay like this.
I almost did not notice that little red between the numbers and thought it would be 115.

To reproduce:

enable this for the view_switcher example
choose View 1 "Simple Button"
choose View 3
repeat from 2. until the ids are >100
choose View 1 or View 4 to get the overlapping ids

cmyr · 2020-04-24T19:30:37Z

yea, I was just working on that. It's kind of tricky; the easiest thing I can do is use z-order painting, so at least we prioritize the topmost widget. I'm also experimenting with only painting if we're currently hot. Not sure what combination of these things makes the most sense...

cmyr · 2020-04-24T20:01:53Z

Okay, I've changed the behaviour to paint using z-order, so children overwrite parents; this should also fix the issue of smaller labels being written over larger labels, because children should always have larger ids than parents. (const ids might violate this, but I don't want to worry about that yet).

In addition, we now only paint the label if the widget is hot.

luleyleo · 2020-04-24T20:34:18Z

The new approach feels better 👍
It does seem like it draws a bit outside the widget tho, which causes visual artifacts as only the widget gets invalidated.

luleyleo · 2020-04-24T20:38:50Z

The invalidation (I guess) also causes problems with the Current view: x label in the view_switcher example. It sometimes does not get a label when hovering or the label is not removed when leaving the label.

cmyr · 2020-04-25T01:48:57Z

right, we're stroking the boundary so it may spill. can adjust that.

about the view_switcher example.... that's weird. My spidey sense says this is an existing bug that we're just seeing for the first time, but I'll dig in a bit more tomorrow.

cmyr · 2020-05-08T14:57:20Z

Okay fixed this, was an issue with how I was handling hot changes.

xStrom · 2020-05-08T17:24:24Z

druid/src/core.rs

+        if debug_layout {
+            self.debug_paint_layout_bounds(&mut inner_ctx, env);
+        }
+        if debug_ids {
+            self.debug_paint_widget_ids(&mut inner_ctx, env);
+        }


The debug_paint_widget_ids method might call debug_paint_layout_bounds, so I think we can optimize it into something like this:

// Check the bool first, because most widgets aren't hot and we won't need to do the map lookup let debug_ids = inner_ctx.is_hot() && env.get(Env::DEBUG_WIDGET_ID); let debug_layout = if debug_ids { // Change debug_paint_widget_ids to return true if it called debug_paint_layout_bounds. // Then we won't need to call it again or even do the env map lookup. !self.debug_paint_widget_ids(&mut inner_ctx, env) && env.get(Env::DEBUG_PAINT) } else { env.get(Env::DEBUG_PAINT) }; if debug_layout { self.debug_paint_layout_bounds(&mut inner_ctx, env); }

okay, I think we can assume that debug_paint_widget_ids always paints the layout so I can simplify this a bit further, overall looks reasonable though.

xStrom

Looks good and useful.

luleyleo

This breaks GTK invalidation. Only some parts get invalidated, buttons for example do not indicate hover, or do so only in one corner. Text field selection does not show up, but while typing text it does work.

~~This only happens when debug_widget_id is active.~~

It happens even without debug_widget_id.

After merging master (which gave me a conflict, wtf?), I tried X11 and it works fine. So this really seems to be GTK specific.

luleyleo · 2020-05-11T18:23:05Z

So, after that merge conflict irritated me, I forcefully reset my local git repo and now the rendering bug is gone 🙈 . But it still does not seem to work entirely, some widgets do not show their id, or have it offset (which could cause some ids to be offscreen)

luleyleo · 2020-05-11T18:27:12Z

After comparing the behavior with X11 (where it works correctly) in the calc example, it seems like ids are drawn to far below and right of the widget they belong to. Also they sometimes do not show up at all, probably because the place where they draw is not being invalidated.

xStrom · 2020-05-11T18:46:39Z

I tested this on my Ubuntu 20.04 just now and both X11 and GTK work just fine. There are no rendering bugs and the ids are drawn in the correct location.

Maybe there's something wrong with your setup @Finnerale? Give cargo clean a try.

luleyleo · 2020-05-11T18:55:37Z

Neither cargo clean nor a complete re-clone of the repo changed it.

luleyleo · 2020-05-11T19:08:57Z

I can even reproduce this on my Ubuntu 20.04 laptop.

cmyr · 2020-05-11T19:31:10Z

hm, so something like this was happening, but I think I've addressed it in an update? is it possible you're on a commit other than 3396138?

luleyleo · 2020-05-11T19:40:07Z

Nope, it is 3396138 . Like I said, I've deleted my local copy of druid and cloned it again, compiled it fresh and the issue still is there, on my Arch desktop and on my Ubuntu 20.04 laptop, both with up to date system and Rust.

cmyr · 2020-05-11T21:25:58Z

that's really strange, because I had an issue similar to what you'd described, but it's now fixed. Can you share a screencap?

This will paint the id of each widget of some subtree in the bottom right of that widget's bounds; useful for debugging various bits of event flow and widget behaviour.

This ensures that the ids of children will not be obscured by the ids of their parents. This also adds caching of the textlayout object.

luleyleo · 2020-05-12T08:05:24Z

So, I've found the reason why only I got it: It happens on Wayland, not on X11.

I've added two videos, the first is on Wayland, the second (same computer) on X11.
wayland.webm
x11.webm

xStrom · 2020-05-12T11:55:26Z

I managed to reproduce the issue with Wayland. This problem has nothing to do with this PR though, the same problem already exists with the anim example. It seems to be related to paint_with_z_index and can be addressed separately from this PR here. That function internally calls self.render_ctx.current_transform() which returns a different value on Wayland.

luleyleo

Interesting, I thought it would have been introduced in this PR because when I tested this PR the last time, it did work. In that case I'm happy to get this merged.

cmyr · 2020-05-12T14:50:58Z

ah that makes sense, I had a similar issue on the coregraphics background where I couldn't trust the current transform returned by the platform, and have to track it separately myself.

cmyr force-pushed the debug-widget-id branch 2 times, most recently from f4d005e to f392919 Compare April 24, 2020 20:08

cmyr mentioned this pull request Apr 24, 2020

Fix incorrect LifeCycleCtx with FocusChanged event #878

Merged

luleyleo added the S-waiting-on-author waits for changes from the submitter label Apr 28, 2020

cmyr force-pushed the debug-widget-id branch from f392919 to 7495cb8 Compare May 8, 2020 14:55

cmyr added S-needs-review waits for review and removed S-waiting-on-author waits for changes from the submitter labels May 8, 2020

xStrom reviewed May 8, 2020

View reviewed changes

cmyr force-pushed the debug-widget-id branch from 6c66bd1 to 3396138 Compare May 11, 2020 16:16

xStrom approved these changes May 11, 2020

View reviewed changes

xStrom removed the S-needs-review waits for review label May 11, 2020

luleyleo requested changes May 11, 2020

View reviewed changes

cmyr added 3 commits May 11, 2020 17:26

Add debug option to display widget ids

cc2178d

This will paint the id of each widget of some subtree in the bottom right of that widget's bounds; useful for debugging various bits of event flow and widget behaviour.

Paint debug widget ids with z-order

d31d9be

This ensures that the ids of children will not be obscured by the ids of their parents. This also adds caching of the textlayout object.

Update changelog

8abd867

cmyr force-pushed the debug-widget-id branch from 3396138 to 8abd867 Compare May 11, 2020 21:28

luleyleo approved these changes May 12, 2020

View reviewed changes

xStrom mentioned this pull request May 12, 2020

GTK: paint_with_z_index is broken #922

Closed

cmyr merged commit 0d56631 into master May 12, 2020

cmyr deleted the debug-widget-id branch May 12, 2020 14:52

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add debug option to display widget ids #876

Add debug option to display widget ids #876

cmyr commented Apr 24, 2020

luleyleo commented Apr 24, 2020

cmyr commented Apr 24, 2020

cmyr commented Apr 24, 2020

luleyleo commented Apr 24, 2020

luleyleo commented Apr 24, 2020

cmyr commented Apr 25, 2020 •

edited

Loading

cmyr commented May 8, 2020

xStrom May 8, 2020

cmyr May 11, 2020

xStrom left a comment

luleyleo left a comment •

edited

Loading

luleyleo commented May 11, 2020

luleyleo commented May 11, 2020 •

edited

Loading

xStrom commented May 11, 2020

luleyleo commented May 11, 2020

luleyleo commented May 11, 2020

cmyr commented May 11, 2020

luleyleo commented May 11, 2020

cmyr commented May 11, 2020

luleyleo commented May 12, 2020

xStrom commented May 12, 2020 •

edited

Loading

luleyleo left a comment

cmyr commented May 12, 2020

Add debug option to display widget ids #876

Add debug option to display widget ids #876

Conversation

cmyr commented Apr 24, 2020

luleyleo commented Apr 24, 2020

cmyr commented Apr 24, 2020

cmyr commented Apr 24, 2020

luleyleo commented Apr 24, 2020

luleyleo commented Apr 24, 2020

cmyr commented Apr 25, 2020 • edited Loading

cmyr commented May 8, 2020

xStrom May 8, 2020

Choose a reason for hiding this comment

cmyr May 11, 2020

Choose a reason for hiding this comment

xStrom left a comment

Choose a reason for hiding this comment

luleyleo left a comment • edited Loading

Choose a reason for hiding this comment

luleyleo commented May 11, 2020

luleyleo commented May 11, 2020 • edited Loading

xStrom commented May 11, 2020

luleyleo commented May 11, 2020

luleyleo commented May 11, 2020

cmyr commented May 11, 2020

luleyleo commented May 11, 2020

cmyr commented May 11, 2020

luleyleo commented May 12, 2020

xStrom commented May 12, 2020 • edited Loading

luleyleo left a comment

Choose a reason for hiding this comment

cmyr commented May 12, 2020

cmyr commented Apr 25, 2020 •

edited

Loading

luleyleo left a comment •

edited

Loading

luleyleo commented May 11, 2020 •

edited

Loading

xStrom commented May 12, 2020 •

edited

Loading