Ren'Py 6.13 Released

[shizuka]

I can reproduce the crash with FBO or Copy RTT

Intel HD Graphics, Core i3/i5/i7 family.
FBO RTT crash on exit with Intel driver 8.10.2361 (RenPy 6.13.0.1603)
FBO RTT crash on exit with Intel driver 8.10.2509 (RenPy 6.13.0.1603)
Copy RTT crash on exit with Intel driver 8.10.2509 (RenPy 6.13.1.1607)

Steps to reproduce:
1. Open Tutorial Game using ANGLE renderer
2. Start game
3. Ctrl to menu showing topic demo Buttons
4. Attempt to close Ren'Py
5. Graphics driver crashes.

Code: Select all

Using shader environment.
Using FBO RTT.
FBO Maximum Texture Size: 8192
Using angle renderer.
FBO Maximum Texture Size: 8192
Performance test:
- Target is 5 frames in 0.333333333333 seconds.
- Frame drawn at 0.000000 seconds.
- Frame drawn at 0.026000 seconds.
- Frame drawn at 0.058000 seconds.
- Frame drawn at 0.075000 seconds.
- Frame drawn at 0.109000 seconds.
- 0.109000 seconds to render 5 frames.
<--no log entry for crash.

Using shader environment.
Using copy RTT.
Using angle renderer.
Performance test:
- Target is 5 frames in 0.333333333333 seconds.
- Frame drawn at 0.000000 seconds.
- Frame drawn at 0.033000 seconds.
- Frame drawn at 0.048000 seconds.
- Frame drawn at 0.065000 seconds.
- Frame drawn at 0.099000 seconds.
- 0.099000 seconds to render 5 frames.
<--no log entry for crash.

[/size]

[shizuka]

Another bug - possibly related to 6.11.2 "The maximum default physical size of the Ren'Py window is now 102 pixels smaller than the height of the screen. This should prevent Ren'Py from creating windows that can't be resized since they are much bigger than the screen."

When starting a 1024x768 visual novel on a 1366x768 screen windowed, interesting things happen:
(1) the visual novel is smaller than the full height of the screen (expected)
(2a) Under OpenGL/DirectX, a customized mouse cursor does not show up past a particular point on the right side of the screen, but things can interact with it. This acts like a clipping border, where the mouse cursor is chopped off at a particular point.
(2b) Under software rendering, the screen is scaled properly initially. Redrawing the area under the mouse uses unscaled images, and creates visual corruption.

This is using an ImageMap for the main menu.

[PyTom]

I just put up another 6.13.1, the result of a review to look for places where the GL/ANGLE path could be screwing up. The most likely possibility was that I didn't disable vertex attributes when switching shaders. So I've added code to do this disabling, and posted that up. Could people who have been having problems update with shift+U again, and let me know if this works for you?

Again, I haven't started on the other bugs yet - I want to get the crashing under control first.

[shizuka]

I still get the driver freeze with 6.13.1.1608; log is the same as 1607.

Maybe just blacklist the "Intel HD Graphics" device. The bug is in the driver after all, and the OpenGL renderer is a good alternative too.

[PyTom]

I still get the driver freeze with 6.13.1.1608; log is the same as 1607.

Maybe just blacklist the "Intel HD Graphics" device. The bug is in the driver after all, and the OpenGL renderer is a good alternative too.

While that's one strategy, I kind of would like to avoid it if possible. For one thing, it's not possible (AFAIK) to get the real device id from inside ANGLE, and even if I could, by the time DirectX is initialized, I've pretty much given up the opportunity to switch to GL. (And vice versa.) At least, I've had problems with that in the past.

Apart from this crash on shutdown, is anyone else having crashing-type problems with 6.13.1.1608? Since I haven't been able to reproduce these problems on my computer, I need feedback from the community if we're finally going to get a stable 6.13 release. (And feedback will help me figure out what to work on next.)

[Aleema]

Apart from this crash on shutdown, is anyone else having crashing-type problems with 6.13.1.1608? Since I haven't been able to reproduce these problems on my computer, I need feedback from the community if we're finally going to get a stable 6.13 release. (And feedback will help me figure out what to work on next.)

If you're interested, my crashing problem with the in-game menu persists with the update.

[PyTom]

If you're interested, my crashing problem with the in-game menu persists with the update.

I'm very interested.

Right now, I'm thinking that my next step will be to create the capability to log all OpenGL calls to a file, and release a version of Ren'Py with this capability. That would let me pinpoint exactly what is causing the problem, in a way that is more reliable that what we have now - which is me just guessing.

[PyTom]

Question to people that have been having problems: Does anyone have a problem with the latest version of Ren'Py when forcing OpenGL? Or is it just the DirectX mode that's problematic?

I'm considering changing the priority of DX and OpenGL, such that we will only use DX if the OpenGL on the system is incapable of running Ren'Py.

[shizuka]

Only having problems when using ANGLE. OpenGL runs just fine.

[PyTom]

Got quite a bit done this weekend, progressing towards 6.13.

I've redone the way Ren'Py picks the display driver to use. Now it will only use DirectX if the OpenGL driver doesn't provide shader support - which pretty much means we only use DX on netbooks, where it seems to make the biggest difference.

I also caught a major bug in the truetype font code that only happened with a negative line spacing. This bug could lead to memory corruption, and hence completely unrelated problems.

I have a strategy for fixing some of the other rendering glitches in newtext, but it's fairly complicated, and so it may take me a few days to make it work.

I'm not going to hold 6.13.1 up for the soft hyphen support, but (unless I run into major problems) I'll add it in a quick 6.13.2, once 6.13.1 is available.

[PyTom]

A pre-release of Ren'Py 6.13.1 is now available. You can get it by:

1) Downloading and unpacking Ren'Py 6.13.0.
2) Hitting shift+U in the launcher.

This attempts to address all bug reports except for one - that one was an oversight I just remembered, and is a very minor one regarding autosave. More specifically:

- Framebuffer objects will not be automatically used.

- It changes the order of the OpenGL and Angle/DX drivers. Ren'Py will now default to using OpenGL on any system that supports shaders. On a windows system that supports shaders only on DirectX, Ren'Py will use DirectX. (This pretty much only applies to older Intel cards, like my netbook's GMA 950.) If DirectX doesn't work, then a fixed-function OpenGL method will be used, and the software renderer is the method of last resort. As with 6.13, users will be warned about fixed-function GL or software rendering.

- A problem involving mouse clipping on scaled displays was fixed.

- Several bugs in font rendering were fixed. Most of these only manifested with a negative line_spacing, but one was bad enough to cause potential memory corruption.

- A new line_overlap_split style property controls where lines are split when using slow text mode. If the bottom of a line is being cut off, you can increase this number from the default of 0.

- Vertical spacing in text blocks is now handled (I think) correctly, when line_spacing is negative. Line bottoms are no longer clipped. There has been one change since 6.12 - in 6.12, when line_spacing was negative, extra space was added to the bottom of a block of text. The space added was (-line_spacing * (number_of_lines - 1)). This space is now gone by default, but can be added back by setting config.broken_line_spacing to True.

- Newlines are allowed in translations.

- Repeating timers now work correctly.

- Several documentation and tutorial game fixes contributed by Koichi Akabee, and one documentation fixe suggested by Soraminako.

Please, please test this pre-release, with your games on the widest range of computers possible. Especially if you had crashes with 6.13.0, please test this. I need to know that Ren'Py will work on as many systems as possible without crashing. I really don't want to make a 6.13.2 under time pressure - paying work is building up - so I want to work with the community to make 6.13.1 the best possible. If you wait until 6.13.1 is out, and then report a bug, it may take a while for me to fix it.

Thank you to everyone who reported bugs in 6.13, and to everyone who will be testing this pre-release.

[mugenjohncel]

Updated Renpy and it fixed the black line issue in the fullscreen but the clipping issue while in windowed mode still remain.

Tested on two machines:
- (A really crappy) Eee PC 1215T
- A standard desktop computer with 6 Gig ram and Palit GeForce GTX 560 2GB (2048MB GDDR5) video card



I swear this is the very same screen based imagemap that is working perfectly fine in the previous renpy and I didn't do anything strange or stupid this time...

"POOF" (Disappears)

[PyTom]

Mugen, are these 1:1 scale images?

Also, could you make the log.txt file available?

Thanks.

[mugenjohncel]

I currently have no access to the previous two machines I tested earlier for at least a few hours from now but here's the log file and the specs of my primary office computer that also displays the same behavior. I don't think this is 1:1 since the screen is 1280 x 1024 while the game is 1280 x 720... this particular computer is using NVIDIA GeForce 9500 GT with drivers up to date...

Here is the log file...

"POOF" (Disappears)

[PyTom]

Also, can you post that gui? I'd like to see if I can trigger the error on one of my computers. I have access to the new pet project forum.