Ren'Py 6.11.0 Public Pre-Release

[Mihara]

One more bug, certain transitions when used in ComposeTransition do not look identical between OpenGL and software mode. Full replication follows.

Code: Select all

init:

    # Images taken from the tutorial game for repeatability.
    
    python:
        transimage = Image("id_dream.png")
        
        transin = ImageDissolve(transimage,1,alpha=True)
        
        testrans = ComposeTransition(Dissolve(1), after=transin)
        testransb = ComposeTransition(transin, after=Dissolve(1))
        
        e = Character('Eileen', color="#c8ffc8")

    image bg a = Solid("#f00")
    image testimage = Image("eileen_concerned.png")

label start:

    scene bg a
    e "This transition works the same in both software mode and OpenGL."
    
    show testimage at truecenter with testrans
    pause
    hide testimage
    
    e "The documentation is not terribly clear on what result should I expect from this one, but it's clearly different between OpenGL and software mode."
    
    show testimage at truecenter with testransb
    pause
    hide testimage

[PyTom]

Mihara >>> Fixed. The GL renderer had the right behavior, while the software renderer was failing to include an alpha channel, leading to unexpected transparency.

[PyTom]

I just uploaded 6.11.0h to the Ren'Py page. The changes are:

- The way we deal with upscaling has been changed. The new approach puts a border around each image. This should prevent us from getting those resizing artifacts introduced in 6.11.0g.
- We consistently use yesno_prompt as the name of the yes/no screen.
- The window size is preserved during a shift+R reload.
- Ren'Py starts when the GL modules fail to load.
- Shift+Y works.
- Several lint problems were fixed.
- The mouse focus tracking has been improved, preventing problems with a hidden mouse when going fullscreen.
- A small performance optimization in GL code.
- in screens, the "group" parameter has been renamed to "style_group", more accurately reflecting it's purpose.
- There's a new If function, that can be used to make an Action conditional.
- F11 toggles fullscreen.
- The drawing area is now constrained using glViewport. This is both a performance improvement, and should hopefully fix Jake's GL clipping problem.
- Dissolves now handle alpha blending properly, in software mode.
- Window show mode now has better logic for picking between NVL and ADV mode.
- We've tweaked when we run the hide event, and documented when the hide event won't run in the ATL documentation

Finally, numbers in the output of lint are now grouped into thousands, using commas. Your lint report will never be the same, unless you have less than 1000 words in your game, in which case it will be exactly the same.

Once again, I thank everyone who submitted bug reports, especially those who submitted game examples to demonstrate problems.

[Mihara]

Umm... Lint doesn't work at all anymore and tracebacks when the 'sound' channel is redefined with renpy.music.register_channel. Replication follows:

Code: Select all

init python:

    renpy.music.register_channel('sound','sfx',loop=False,stop_on_mute=False, tight=False,file_prefix="sound/", file_suffix=".ogg")

    e = Character('Eileen', color="#c8ffc8")

init:    
    
    image bg a = Solid("#f00")

label start:
    
    play sound "something"

    scene bg a
    e "Lint shouldn't crash on this, it should report that 'something' is not loadable."

Traceback follows:

Code: Select all

Ren'Py 6.11.0h lint report, generated at: Sat Aug 07 18:01:40 2010
  File "c:\Gamedrive\renpy-6.11.0\renpy\bootstrap.py", line 252, in bootstrap
  File "c:\Gamedrive\renpy-6.11.0\renpy\main.py", line 289, in main
  File "c:\Gamedrive\renpy-6.11.0\renpy\lint.py", line 478, in lint
  File "c:\Gamedrive\renpy-6.11.0\renpy\lint.py", line 230, in check_user
  File "c:\Gamedrive\renpy-6.11.0\renpy\ast.py", line 1345, in call
  File "c:\Gamedrive\renpy-6.11.0\renpy\statements.py", line 92, in call
  File "renpy-6.11.0/common/00statements.rpy", line 264, in lint_play_sound
NameError: global name 'lint_play_music' is not defined

After initialization, but before game start.
 - python at line 264 of renpy-6.11.0/common/00statements.rpy.

[Mihara]

And another one.

Shift+R now always causes RenPy to restart in software mode. Whether this is due to Shift still being held down or because the test doesn't run correctly when restarting, I cannot confirm, but all attempts to let go of the shift key early enough have so far failed.

[Mihara]

And yet another one.

Changing window size along the horizontal axis to produce vertical letterboxing (such as when running a 16:9 project on a 4:3 monitor) produces, alternatively, aspect ratio distortion or no aspect ratio adjustments whatsoever. Looks like it's related to whatever you did with glViewport.

Pictures and opengl.txt included.

[PyTom]

And another one.

Shift+R now always causes RenPy to restart in software mode. Whether this is due to Shift still being held down or because the test doesn't run correctly when restarting, I cannot confirm, but all attempts to let go of the shift key early enough have so far failed.

Can you give me the opengl.txt? It should tell me which of these it is.

[Mihara]

Can you give me the opengl.txt? It should tell me which of these it is.

Here. Says it's because shift is held down.

[backansi]

I can resize the window but can't click the full-screen button on the title bar... (OS: Windows )
IIRC it worked well in previous version. Is it prevented on purpose?

[PyTom]

Mihara: Okay, fixed all three.

- I had actually caught the lint problem, but for some reason didn't upload my fixed version.
- Vertical scaling was a math problem.
- I've modified OpenGL to remember the status of shift at startup, and not check it again.

Please let me know when you're done testing, and I'll cut a new release if you don't find anything else.

backansi: Yes, that's on purpose. With the maximize button, it's too easy to get the window to grow until the start orb covers it - at which point performance drops from ~300 fps to ~25.

A longer-term approach for this might be to wait for Google's ANGLE project to mature a bit, and then switch the windows version to use DirectX via ANGLE. But that won't be ready for this release.

[Mihara]

Well, here's two more bugs:

1. Software mode seems to assume it's running on a 4:3 monitor, which is not necessarily the case.

config.screen_width = 768, config.screen_height = 480. (i.e. a 16:10 project) goes fullscreen on a large 16:10 monitor in software mode. Expected result: Actual fullscreen mode switching or some kind or an error message because the mode is not supported.
Actual result: It seems to switch to 800x600 and letterbox the output instead.

I think that one can wait.

2. Screens defined with "modal True" prevent developer mode shortcuts from running, too, making them hard to iteratively adjust.
EDIT: All keyboard shortcuts actually, "F" to go fullscreen included.

[PyTom]

1) Is unfixable in software mode - you have to use GL mode to get correct behavior here.

2) I will look into.

[Mihara]

Ok, and one that is NOT a bug, but odd.

My problem with playing video on Eee 901/Linux that you have thought system specific mysteriously went away without a comment -- if you did anything, it would be a good idea to find out what to prevent regression.

I think that's it for the moment, I don't think I can find any more obvious problems with 6.11h, so it's back to development for me until I trip over an obscure one. Next up is getting screens to do what I want...

[PyTom]

It could have been the viewport change, maybe - but I don't think so. Frankly, I have no idea what caused it, and no idea what made it go away.

I've fixed all of Mihara's bugs. I would like Jake and Spiky to try out this release, to make sure it solved the Mac clipping and Linux focus problems, respectively. Then I'll release these fixes for some testing, and make the release.

[Spiky Caterpillar]

The Linux focus bug appears fixed by 6.11.0h.

With the maximize button, it's too easy to get the window to grow until the start orb covers it - at which point performance drops from ~300 fps to ~25.

How odd. I vaguely wonder if it might make sense to have maximized windowed-mode Seven switch to software mode?

(I don't have seven, though, so it doesn't affect me directly - and removing the maximize button for seven users will work almost as well and should be less likely to have unexpected side effects.)