Gtk4 port #53

jwahlstrand · 2023-07-02T02:36:44Z

This is a port to use Gtk4.jl instead of Gtk.jl.

The primary benefit of Gtk4.jl is that it's based on the most actively maintained version of GTK. My impression is that it works better on Windows and MacOS than version 3, but I should note that there are a few annoying bugs on Windows. Other advantages: Gtk4.jl doesn't cause horrendous REPL lag on Windows, it works well with PackageCompiler, at least in my limited experience, it allows GL-based rendering on all three platforms (including Makie via Gtk4Makie), and it supports the vast majority of widgets and methods in the underlying C libraries through GObject introspection.

I have been using Gtk4.jl in my own projects for a while now, as has @tknopp. Documentation needs work and that is a priority, but the current docs are about the same level as Gtk.jl.

Gtk4 only rounds when the value is shown, which is not the default

jwahlstrand · 2023-08-06T00:26:29Z

I think this is ready to go now -- I don't see how to simulate combinations of mouse clicks and keyboard modifiers, which are used in tests and precompiling in the GTK3 version. Maybe doing so would involve GtkEventControllerLegacy (which I really don't want to go near) and even that may not work. Anyway, these types of events work fine for me (in the "imageviewer.jl" example and the corresponding ports of ProfileView and ImageView).

timholy · 2023-08-09T19:49:10Z

I will hope to get to this over the weekend. The line-change-count is intimidating but I'm sure it was far harder to execute than it will be to review! I would want to spend a bit of time seeing if I can figure out any way to restore automation of those tests, but if you haven't found anything I suspect there's very little chance that I'll discover anything. It does seem weird, though; how does Gtk4 do its own internal testing if these things can't be automated?

codecov-commenter · 2023-08-09T23:53:41Z

Codecov Report

Merging #53 (9db5582) into master (dca6f5f) will increase coverage by 2.86%.
The diff coverage is 88.09%.

@@            Coverage Diff             @@
##           master      #53      +/-   ##
==========================================
+ Coverage   91.62%   94.49%   +2.86%     
==========================================
  Files           6        6              
  Lines        1098     1089       -9     
==========================================
+ Hits         1006     1029      +23     
+ Misses         92       60      -32

Files Changed	Coverage Δ
src/GtkObservables.jl	`89.09% <61.53%> (-5.74%)`	⬇️
src/widgets.jl	`92.90% <80.95%> (+0.34%)`	⬆️
src/graphics_interaction.jl	`94.73% <91.89%> (-0.54%)`	⬇️
src/extrawidgets.jl	`95.48% <92.85%> (-0.88%)`	⬇️
src/precompile.jl	`99.13% <97.61%> (+31.03%)`	⬆️
src/rubberband.jl	`96.59% <100.00%> (ø)`

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

jwahlstrand · 2023-08-10T01:01:48Z

Thanks! It looks like there is a problem with tests on nightly related to timers, which I saw mention of in the precompile code. I'll look into it.

Yeah, the most difficult part of this port was adapting to the changes to event handling, and those are unfortunately the parts that aren't currently being tested. I'll take a closer look at how to simulate events. You're right, they must do this somehow within GTK and it would be worth examining how they do testing within GTK. I'm waist deep in this so I think it's up to me to get this working.

I tried not to make gratuitous edits, but I did make a few changes that are not directly related to the port to Gtk4.jl: I tried to fix some of the doctests, which weren't working (and oops, still aren't I guess), and updated some versions in CI that lead to warnings. If it would make your task easier I could split those into separate PR's, but it's not going to reduce this PR by much.

timholy · 2023-08-10T17:28:32Z

Make life easy on yourself and leave them here.

timholy

Other than the issue about testing discussed earlier, I see nothing but goodness here. Merge when ready.

timholy · 2023-08-13T21:15:02Z

docs/src/controls.md

@@ -2,11 +2,10 @@

 Let's create a `slider` object:
 ```jldoctest demo1
-julia> using Gtk.ShortNames, GtkObservables


Doesn't have to be fixed in this PR, but we'd want a path towards resolving doc errors too.

Agreed -- it looks like we need to do some pattern matching to get them to pass. I will submit a new PR once I figure out how to do that.

Yeah, definitely don't worry about that for this PR, unless you know of something that will make that really hard. One of us can tackle that later.

timholy · 2023-08-13T21:17:59Z

docs/src/drawing.md

@@ -34,7 +33,7 @@ We're going to set this up so that a new line is started when the user
 clicks with the left mouse button; when the user releases the mouse
 button, the line is finished and added to a list of previously-drawn
 lines. Consequently, we need a place to store user data. We'll use
-Signals, so that our Canvas will be notified when there is new
+Observables, so that our Canvas will be notified when there is new


timholy · 2023-08-13T21:21:06Z

src/GtkObservables.jl


 # Re-exports
 export set_coordinates, BoundingBox, SHIFT, CONTROL, MOD1, UP, DOWN, LEFT, RIGHT,
-       BUTTON_PRESS, DOUBLE_BUTTON_PRESS, destroy


What's the replacement for double-click?

In GTK4 it's one of the arguments in the signal handler.

timholy · 2023-08-13T21:21:59Z

src/extrawidgets.jl

+    step_back = button(; widget=builder["step_back$id"]::Gtk4.GtkButton)
+    stop = button(; widget=builder["stop$id"]::Gtk4.GtkButton)
+    step_forward = button(; widget=builder["step_forward$id"]::Gtk4.GtkButton)
+    play_forward = button(; widget=builder["play_forward$id"]::Gtk4.GtkButton)


Much nicer!

timholy · 2023-08-13T21:22:50Z

src/extrawidgets.jl

-    destroy(p.entry)
-    destroy(p.scale)
-    destroy(frame(p))
-end


I take it these are handled automatically now?

Yeah, there's no destroy for widgets now, just for windows. If you destroy windows typically all the widgets are destroyed too.

jwahlstrand · 2023-08-15T21:00:42Z

Ugh, I just tried capturing the modifier state in a way that could be tested and precompiled, only to run into a GTK4 bug: https://gitlab.gnome.org/GNOME/gtk/-/issues/5139

I didn't find any tests for this stuff inside the C library, but maybe I didn't look in the right place. I will keep looking but I'm nearing my limit.

timholy · 2023-08-16T16:24:04Z

I think we can proceed here anyway if it's not testable and you think it's an improved experience. (I should say I haven't tested myself.) At some point, if there's a gtk dev forum maybe one of us could ask them how they do testing?

timholy · 2023-08-16T16:25:14Z

In other words: if you're to the point of wanting to click merge, just do so. If you still want to play with it for a bit more, that's fine too.

This introduces a way to override the modifier state for the purposes of testing. It's better than nothing.

jwahlstrand · 2023-08-16T17:43:26Z

I think we can proceed here anyway if it's not testable and you think it's an improved experience. (I should say I haven't tested myself.) At some point, if there's a gtk dev forum maybe one of us could ask them how they do testing?

OK, I got something to work -- it doesn't generate real GTK events but it exercises the Julia code. I just want to add precompiles using that too, then I'll merge.

timholy · 2023-08-19T08:08:13Z

Hooray!

Let's get this out there. I presume this should be 2.0.0? E.g., changes like DOUBLE_BUTTON_PRESS and people who built Gtk3 GUIs with glade, etc, are going to have to make some changes?

jwahlstrand · 2023-08-19T12:23:07Z

Yup, this should definitely be 2.0! Go for it, otherwise I'll have time tomorrow evening (assuming I have that access level).

timholy · 2023-08-19T13:11:28Z

You do, but I'll do it now.

timholy · 2023-08-19T13:14:07Z

JuliaRegistries/General#89925

Thanks again, I'm moving my whole lab but when the dust settles I'm excited to start playing with this!

jwahlstrand added 12 commits July 1, 2023 15:27

use Gtk4

e74678a

support "focus-leave" event for textbox, restore some tests

fcc8b74

fix some tests

8058b4a

add XML file (formerly player.glade)

8bc3870

restore commented out code, fix a crash in slider

aa1109b

fix slider

2925bda

Gtk4 only rounds when the value is shown, which is not the default

fix zoom-scroll at mouse position

21ab3f3

work on precompile code

6697a77

fix a few doctest problems, remove 1.3 from CI

73bf4b5

fix the canvas precompiles that I know how to fix

371a715

don't open windows during precompilation

340dad9

bump Gtk4 compat

7475499

timholy reviewed Aug 13, 2023

View reviewed changes

implement rest of tests

6bba9e6

This introduces a way to override the modifier state for the purposes of testing. It's better than nothing.

add precompile code for events with modifiers

9db5582

jwahlstrand merged commit ae4802b into JuliaGizmos:master Aug 18, 2023

jwahlstrand deleted the gtk4 branch August 19, 2023 12:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Gtk4 port #53

Gtk4 port #53

jwahlstrand commented Jul 2, 2023 •

edited

Loading

jwahlstrand commented Aug 6, 2023

timholy commented Aug 9, 2023

codecov-commenter commented Aug 9, 2023 •

edited

Loading

jwahlstrand commented Aug 10, 2023

timholy commented Aug 10, 2023

timholy left a comment

timholy Aug 13, 2023

jwahlstrand Aug 15, 2023

timholy Aug 16, 2023

timholy Aug 13, 2023

timholy Aug 13, 2023

jwahlstrand Aug 15, 2023

timholy Aug 13, 2023

timholy Aug 13, 2023

jwahlstrand Aug 15, 2023

jwahlstrand commented Aug 15, 2023

timholy commented Aug 16, 2023 •

edited

Loading

timholy commented Aug 16, 2023

jwahlstrand commented Aug 16, 2023

timholy commented Aug 19, 2023

jwahlstrand commented Aug 19, 2023

timholy commented Aug 19, 2023

timholy commented Aug 19, 2023

Gtk4 port #53

Gtk4 port #53

Conversation

jwahlstrand commented Jul 2, 2023 • edited Loading

jwahlstrand commented Aug 6, 2023

timholy commented Aug 9, 2023

codecov-commenter commented Aug 9, 2023 • edited Loading

Codecov Report

jwahlstrand commented Aug 10, 2023

timholy commented Aug 10, 2023

timholy left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jwahlstrand commented Aug 15, 2023

timholy commented Aug 16, 2023 • edited Loading

timholy commented Aug 16, 2023

jwahlstrand commented Aug 16, 2023

timholy commented Aug 19, 2023

jwahlstrand commented Aug 19, 2023

timholy commented Aug 19, 2023

timholy commented Aug 19, 2023

jwahlstrand commented Jul 2, 2023 •

edited

Loading

codecov-commenter commented Aug 9, 2023 •

edited

Loading

timholy commented Aug 16, 2023 •

edited

Loading