ggez-0.6 update #50

PSteinhaus · 2021-09-04T08:59:43Z

Ok, this PR is going to be a big one, sry.
But it's not as bad as it seems, as most of the long code is just copy-pasted examples from ggez.

I basically changed the API to conform to ggez-0.6 and then added example after example.
For each example I implemented (i.e. mostly copied from ggez) the functionality that was necessary to run it, thereby pushing good-web-game closer towards being a complete ggez implementation.

If you go through the changes you can see that I also changed some things here and there (for example removing InputHandler from KeyboardContext), but I tried not to touch any sensitive parts like the filesystem, where opinions on how to move forward may differ. So I tried to stick to things like bumping lyon, etc.

I also ran into some nasty problems, which I wasn't able to fix on my own until now.
The most prominent one being the one discussed in #40, that drawing things like text or meshes sometimes leads to a STATUS_ACCESS_VIOLATION, which can sometimes be worked around by adding even more draw calls for some reason, but sometimes can't... (see the many TODOs)

Other problems include things like color being off in the colorspace example, not being able to render to canvases sometimes and miniquad::EventHandlerFree::key_down_event always handing us repeat values of false, even when the key was obviously repeated (this last one should be easy enough to workaround, but I thought I should mention it in case this bug is still present in miniquad; maybe updating to a newer version of miniquad might fix it, haven't checked yet.

Now the question is how we want to advance. We could either keep this PR open and work on my fork, or we could merge it right away and work on it right here (meaning using the issue tracker of this repo instead of my fork). I'd prefer the latter option, but I could of course understand if you didn't want to merge this PR before being 100% content with it.

EDIT: I just saw that #49 also features some fixes, including some that this PR doesn't, so I guess it might be better to merge that one first, perhaps.

added some of the ggez-0.6 examples; expanded functionality (implemented more of ggez-0.6)

PSteinhaus · 2021-09-04T09:15:11Z

Also, I didn't use the latest version of miniquad for creating this PR, but instead sticked to the one used currently.
This might be something to change before merging the PR, maybe.

not-fl3 · 2021-09-04T14:48:16Z

Wow, very impressive work!

The most prominent one being the one discussed in #40, that drawing things like text or meshes sometimes leads to a STATUS_ACCESS_VIOLATION

We had similar problem in macroquad, and doing this(moving delete after SwapBuffers) helped:
#40 (comment).

maybe updating to a newer version of miniquad might fix it, haven't checked yet.

Miniquad used here is really old (0.3-alpha is a recommended version to use, it moved quite a lot from 0.2).

Now the question is how we want to advance. We could either keep this PR open and work on my fork, or we could merge it right away and work on it right here (meaning using the issue tracker of this repo instead of my fork). I'd prefer the latter option, but I could of course understand if you didn't want to merge this PR before being 100% content with it.

I am not actively maintaining good-web-game anymore, so, if you want, I can give you admin rights here on this repo or even transfer you the ownership.

not-fl3 · 2021-09-04T14:51:26Z

Cargo.toml

@@ -14,10 +14,16 @@ cgmath = { version = "0.17", features = ["mint"] }
 miniquad_text_rusttype = { git = "https://github.com/not-fl3/miniquad_text_rusttype" }
 miniquad = "=0.2.55"


miniquad 0.2 is seriously outdated, there are lots, lots of bugfixes in 0.3.0-alpha.37

ok, I'll try to bump it to the latest version of miniquad then ;)

PSteinhaus · 2021-09-04T17:20:08Z

Wow, very impressive work!

Thank you ^^ looking at the line count I feel somewhat impressed myself, but it's mostly copies of the wonderful code made by all the wonderful contributors of ggez, really.

The most prominent one being the one discussed in #40, that drawing things like text or meshes sometimes leads to a STATUS_ACCESS_VIOLATION

We had similar problem in macroquad, and doing this(moving delete after SwapBuffers) helped:
#40 (comment).

Hmm, interesting. I haven't investigated deeply, but I couldn't find any references to glDeleteBuffers in good-web-game, or the surrounding libraries, but I assume it happens as part of ctx.quad_ctx.end_render_pass()?

Looking into this I just realized present() is nothing more than a dummy in good-web-game 😅 I really gotta look into miniquad to find out more about how it handles all of this.

Now the question is how we want to advance. We could either keep this PR open and work on my fork, or we could merge it right away and work on it right here (meaning using the issue tracker of this repo instead of my fork). I'd prefer the latter option, but I could of course understand if you didn't want to merge this PR before being 100% content with it.

I am not actively maintaining good-web-game anymore, so, if you want, I can give you admin rights here on this repo or even transfer you the ownership.

Wow! Thank you for that offer! I'd gladly accept admin rights. It'd make working on this much easier, thank you.
There are different things where I'd want to, at least, ask you for your opinion, both for reasons of experience and original ownership, still.

not-fl3 · 2021-09-04T23:01:45Z

Hmm, interesting. I haven't investigated deeply, but I couldn't find any references to glDeleteBuffers in good-web-game, or the surrounding libraries, but I assume it happens as part of ctx.quad_ctx.end_render_pass()?

IIRC the problem occurs when you

create buffer
draw buffer
delete buffer
do a commit_frame or just end the draw handler

ggez do buffers reallocation all the time, so it happens quite often. But I did not really research whats going on after the error got fixed in macroquad.

Wow! Thank you for that offer! I'd gladly accept admin rights. It'd make working on this much easier, thank you.
There are different things where I'd want to, at least, ask you for your opinion, both for reasons of experience and original ownership, still.

Done, you are an admin (or will be after following this link: https://github.com/not-fl3/good-web-game/invitations
) :)

PSteinhaus · 2021-09-05T17:47:49Z

Ok, updating to miniquad 0.3.0-alpha.37 alone didn't fix anything AFAIK, too bad.
In fact, the astroblasto example is now somewhat broken (but that's probably just due to outdated RNGs and should be easy to fix, hopefully).

IIRC the problem occurs when you
* create buffer

* draw buffer

* delete buffer

* do a `commit_frame` or just end the `draw` handler
ggez do buffers reallocation all the time, so it happens quite often. But I did not really research whats going on after the error got fixed in macroquad.

Hmm, is there a thread on this in your issue tracker? (searched a bit already, but couldn't find one)
Where/when does miniquad swap its buffers?
I have a few questions regarding miniquad in general, I guess I'll take a look into the discord server.

Done, you are an admin (or will be after following this link: https://github.com/not-fl3/good-web-game/invitations
) :)

Thx! ❤️

not-fl3 · 2021-09-06T03:44:02Z

Hmm, is there a thread on this in your issue tracker? (searched a bit already, but couldn't find one)

#40 (comment)

This is the minimum repro and an explanation on why it crashes

Where/when does miniquad swap its buffers?

It depends on the platform. On WebGL there is no swapbuffer whatsoever: https://stackoverflow.com/questions/41018445/where-is-the-swapbuffer-opengl-call-in-webgl/41045668#41045668

On windows, where the crash persists, miniquad calls glSwapBuffer right after a frame callback

PSteinhaus · 2021-09-06T10:03:26Z

Hmm, is there a thread on this in your issue tracker? (searched a bit already, but couldn't find one)

#40 (comment)

This is the minimum repro and an explanation on why it crashes

Of course. Sorry, I should have been more precise. I meant: is there a thread on the macroquad issue tracker that I missed?
I'm not that experienced in OpenGL yet and it would be interesting for me to see how you fixed it in macroquad.
I guess I should take a closer look at macroquad in general.

ported canvas flipping fix from ggez to here (together with other things);

removed now unnecessary mod `rand`;

…ng buffer deletion two frames instead of one; added blend modes; changed start function and configuration;

not-fl3 · 2021-09-10T23:03:12Z

src/lib.rs

@@ -136,18 +139,22 @@ impl<E: std::error::Error> miniquad::EventHandlerFree for EventHandlerWrapper<E>
    }
 }

-pub fn start<F, E>(conf: conf::Conf, f: F) -> GameResult
+pub fn start<F, E>(quad_conf: miniquad::conf::Conf, conf: conf::Conf, f: F) -> GameResult


I do not recommend using miniquad's Conf in public API.

Cache, for example, is gwg-only thing and should be removed from miniquad's Conf.

(cache in miniquad is a legacy from times when miniquad was not able to load files by url)

Ah, I didn't know that. Thx for explaining.
Then I'll put cache back into the good-web-game Conf. It fits better anyway.
I'll probably just add all the miniquad fields to it as well and then simply create a miniquad Conf to hand over when starting the application.

added some missing graphics functions; updated tar file; made canvas flipping slightly more efficient; refactored setting blend modes a little;

started work on a MeshBatch implementation;

added appropriate alpha blending to all pipelines; cleaned up some examples; enabled filters on textures used in meshes; added meshbatch pipeline (though it's not used yet);

PSteinhaus · 2021-09-13T15:38:49Z

Ok, I'm happy with how this turned out :)
All the important stuff from ggez is now included, with one exception: using your own shaders.
I'll not include this feature in this PR, as I'm probably not the right one to write it. I leave it to someone more experienced in the field (i.e. shaders in miniquad).

added default filter functions;

updated documentation; added functions for grabbing/hiding the cursor and setting cursor type; removed Loading from conf, as it currently doesn't to anything;

erlend-sh · 2021-09-14T12:43:11Z

🎉 👏

About page should probably be updated:

[DEPRECATED] A minimal 2d game framework for wasm.

PSteinhaus · 2021-09-14T17:06:26Z

🎉 👏

About page should probably be updated:

[DEPRECATED] A minimal 2d game framework for wasm.

Yup, couldn't do that before getting ownership though ;)

changed the API to conform to ggez-0.6;

a0ba78d

added some of the ggez-0.6 examples; expanded functionality (implemented more of ggez-0.6)

PSteinhaus mentioned this pull request Sep 4, 2021

Discussion on replacing gfx-rs ggez/ggez#962

Closed

not-fl3 reviewed Sep 4, 2021

View reviewed changes

updated to miniquad 0.3.0-alpha.37

b73343b

PSteinhaus added 7 commits September 8, 2021 11:24

Merge branch 'not-fl3:master' into master

54cf915

somewhat fixed #40 (but not completely yet);

45b93eb

ported canvas flipping fix from ggez to here (together with other things);

used quad-rand for rng needs in the examples;

39f5a9b

removed now unnecessary mod `rand`;

fixed color behavior according to the colorspace example;

3d7835f

fixed rng usage in bunnymark

d563c44

changed text implementation (back) to current ggez one (glyph_brush)

7ec6ce1

fixed the buffer deletion bug by applying a thicker band-aid: deferri…

7a4cd1a

…ng buffer deletion two frames instead of one; added blend modes; changed start function and configuration;

not-fl3 reviewed Sep 10, 2021

View reviewed changes

PSteinhaus added 5 commits September 12, 2021 12:51

adapted configuration to no longer expose miniquad::conf::Conf;

e371076

added some missing graphics functions; updated tar file; made canvas flipping slightly more efficient; refactored setting blend modes a little;

fixed the mesh shader to respect the DrawParam::src parameter;

b26d840

started work on a MeshBatch implementation;

removed unnecessary blend mode field from SpriteBatch;

5e7b4be

added appropriate alpha blending to all pipelines; cleaned up some examples; enabled filters on textures used in meshes; added meshbatch pipeline (though it's not used yet);

implemented MeshBatch

90a7a46

rustfmt

aa30db2

This was referenced Sep 13, 2021

The snake example segfaults #40

Closed

Support MeshBatch from ggez #31

Closed

PSteinhaus added 2 commits September 14, 2021 00:12

updated README;

37b05b6

added default filter functions;

updated README;

e58a456

updated documentation; added functions for grabbing/hiding the cursor and setting cursor type; removed Loading from conf, as it currently doesn't to anything;

PSteinhaus merged commit ce64fd4 into ggez:master Sep 14, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ggez-0.6 update #50

ggez-0.6 update #50

PSteinhaus commented Sep 4, 2021 •

edited

PSteinhaus commented Sep 4, 2021 •

edited

not-fl3 commented Sep 4, 2021

not-fl3 Sep 4, 2021

PSteinhaus Sep 4, 2021

PSteinhaus commented Sep 4, 2021

not-fl3 commented Sep 4, 2021

PSteinhaus commented Sep 5, 2021 •

edited

not-fl3 commented Sep 6, 2021 •

edited

PSteinhaus commented Sep 6, 2021

not-fl3 Sep 10, 2021

not-fl3 Sep 10, 2021

PSteinhaus Sep 11, 2021

PSteinhaus commented Sep 13, 2021

erlend-sh commented Sep 14, 2021

PSteinhaus commented Sep 14, 2021

		@@ -14,10 +14,16 @@ cgmath = { version = "0.17", features = ["mint"] }
		miniquad_text_rusttype = { git = "https://github.com/not-fl3/miniquad_text_rusttype" }
		miniquad = "=0.2.55"

ggez-0.6 update #50

ggez-0.6 update #50

Conversation

PSteinhaus commented Sep 4, 2021 • edited

PSteinhaus commented Sep 4, 2021 • edited

not-fl3 commented Sep 4, 2021

not-fl3 Sep 4, 2021

Choose a reason for hiding this comment

PSteinhaus Sep 4, 2021

Choose a reason for hiding this comment

PSteinhaus commented Sep 4, 2021

not-fl3 commented Sep 4, 2021

PSteinhaus commented Sep 5, 2021 • edited

not-fl3 commented Sep 6, 2021 • edited

PSteinhaus commented Sep 6, 2021

not-fl3 Sep 10, 2021

Choose a reason for hiding this comment

not-fl3 Sep 10, 2021

Choose a reason for hiding this comment

PSteinhaus Sep 11, 2021

Choose a reason for hiding this comment

PSteinhaus commented Sep 13, 2021

erlend-sh commented Sep 14, 2021

PSteinhaus commented Sep 14, 2021

PSteinhaus commented Sep 4, 2021 •

edited

PSteinhaus commented Sep 4, 2021 •

edited

PSteinhaus commented Sep 5, 2021 •

edited

not-fl3 commented Sep 6, 2021 •

edited