Reborn return to castle Wolfenstein

@walkero

Thanks, but can you please test again using the *release* kernel. I asked for that in my reply above because the debug version changes the memory layout and it might be an annoying memory trashing bug.

Also, what's the kernel version?

Hans

@Hans
Sorry I missed your request for the kernel.
I use Kernel v53.98 which is the latest beta one.

Below you can find the crashlog with the non debug kernel
http://os4coding.net/node/2316

@walkero

Thanks. It's consistent with your other new logs. However, the new logs crash in a totally different way to the previous ones (ISI instead of DSI).

The whole thing is even more confusing now. I wonder if this might be a clue (from one of your newer logs):

[_impl_ObtainSemaphore:108] *** WARNING: "Semaphore might be uninitialized", called from 0x0184BDDC

Hans

@walkero


Could you try executing the following from within your SYS:kickstart directory:
addr2line -e kernel.debug -j .text 0x30ac

Thomas said it might be a list function like AddHead(), Remove(), etc. Running the command above may help identify what it is.


Quoting Thomas directly:
If it's really a list function, then it's going to be difficult to trace. The most likely candidate for a cause is a buffer overflow.

From the rest of the log posted, I would say the list it was manipulating was the task list :| Also, the fact that you get frequent "suspending already suspended tasks" shows that there is something wrong with the task list.

What I found to be a frequent problem with corrupted task lists (apart from buffer overflows or general memory corruption) is the execution of system functions that cause preemption in a context that shouldn't be preempted. Good example is an interrupt handler. Dunno if RtcW uses any, but maybe a callback for sound or joystick...

Hans

Might as well pass this one on too. It relates to walkero's last crash log:
> Temporary stack trace:
> #0: 0x12141350
> #1: 0x7F549EDC
> #2: 0x7F549F4C
> #3: 0x7F56BEE4
> #4: 0x7F55CA50

> Crashed process: reborn-sp (0x6A464E00)
> 0: 6FFAE000 6A6568C0 00000002 4DE1ECE8 00000000 6A6567B8 00000002 6A462184
> 8: 6386A610 7F4097C8 0000001C 12141352 0000015C 60CA0CAC 33504449 60C99724
> 16: 6A656A5C 6A656A40 43504449 00000000 6A656B48 6A656A38 638F0A28 63969BA0
> 24: 00000000 00000001 00000000 00000000 00000003 60879534 60C9C660 60879000
> CR: 39955593 XER: A000007E CTR: 12141352 LR: 7F4097D8

The crash address is basically an address-corrected version of CTR, which would imply that in this case, the faulty address was jumped to via a CTR jump. To the best of my knowledge, the compiler doesn't create code like that, but it's the default method for jumping into an interface function, which would hint at either a defective interface, or a pointer to a not-so-much-an-interface was used as an interface (if you get what I mean)

Really hope these clues will help track down the root cause.

Hans

@Hans

Quote:

To do that I must boot with the debug kernel or not?

@walkero

Quote:
It doesn't matter, so long as you execute the command from within your kickstart drawer.

Hans

@Hans
AmigaOS:Kickstart> addr2line -e kernel.debug -j .text 0x30ac
??:0

@Hans

I would think you needed a debug build (ie built with -gstabs unstrippped) of the kernel to get meaningful results from addr2line. That'a not what kernel.debug is it's a kernel with extra debug code.

well finally got the HD version to work tonight ! Turns out it was the pakHD02.pk3 file that was corrupt and a re-download solved the problem....game looks and runs like a beauty on my X1000

donation sent over to our man HUNO

@broadblues

Yes, but it was Thomas' suggestion, so I hoped that him suggesting it meant that the debug kernel was unstripped. Obviously not.

Hans

@all
I tried on my x5k with all latest ogles2, warp3dnova and radeonhd.

With default settings it crash for me as for everyone else when i start to load level, and there is what i got on serial:




Reducing those optiosn from High to Medium make game pass init screen , and loading till first videos.

Geometric Detail
Character Textures
General Texture

But still crashes.

And only when i set those 3 options to the Low, i can play in game.

That all remind me what issues i noticed few days ago in Jedi, when i set textures to High, and it lockup, while there is still 1gb of free system memory, and only 700 was in use. And crash was in RadeonHD module, but that mean nothing if memory trashed..

@Huno

I found some new issue with RCTW Reborn. Well, more precissely with EGL_Wrap.

For some reassons you put in prefs/env-archive special GL4ES settigns, like noone else but only EGL_Wrap will use them :)

So they stay there forver, which in turn mean that all the other GL4ES games/apps, will use those environemnts you set just for EGL_Wrap all the time automatically.

In my case, few GL4ES games just looks wrong by colors because of this, so i loose a hour before understand what going on :)

What i mean, is that saving LIBGL environments just for EGL_Wrap should't be done in the main system env: of course, as it cause issues with all other GL4ES stuff.

It can be instead in the game's directories, or dunno, maybe inside of EGL_Wrap code, or alternatively envs can be called EGL_WRAP_LIBGL_ and inside of EGL_Wrap make a redefine of GL4ES ones to those ones , so in ENV: and ENV_ARC: will be those with EGL_WRAP_ at begining, and other GL4ES stuff will not use them.

Because as it now, everyone who will release anything with GL4ES, should note users that they should clean LIBGL_* entries in the system:prefs/env-archive because EGL_Wrap save special environments for their own needs, which in turn will have impact on everything else done via GL4ES :)

Plz fix it, i will happy betatest fixed version if you need tester.

Thanks !

@kas1e
Hi
Yes you're right this can impact other games or applications that take advantage of gl4es. I apologize for the inconvenience caused by studying a solution internally not to make unnecessary excesses on other work
Thank you very much for these feedbacks that help me move forward in my EGL dev
Yes and no problems for betatest
Thanks
HunoPPC

@Huno
What i also noticed, is that in the Level3 , there is major-strange slowdown of the game to just 5-8fps (and that with low textures settings). Dunno to what component that related of course..

Its very visibly when you come to the funicular, and watch on it. Also you can't enter to the room at right of the funicular, you can see soldier, but you can't kill him , and can't come to the room, while door is open. Just like some bug.


Also noticed that when you break a barrel, those parts of barrel which drops on floor have no textures (white ones). Or maybe have, but didn't rendered correctly.

ps. Can you also rebuild RTCW Reborn with latest gl4es (at least with Mar 31, 2019) : ptitSeb add there some new environment by which possible to dump shaders of different types (vertes, fragments, both, before shader conversion, after, etc, etc). So it will be easy to see what shaders in use when rendering bugs happens.

@kas1e
The preferences are now in ENVARC: EGL_Wrap and not at the root so more conflicts. By cons it is not yet functional in the bookstore that I made available because I work on it now. As for the level 3 actually it is very slow, the map has too many 3D objects and PC version rallenti aussimais it is more obvious on the amigaos4, should I see if by modifying some small thing on the map if I can win FPS.
I will test this .
Download the executable and the lib on the ftp and ask me if you want other things.
HunoPPC

@Huno
Checked v8 of egl_wrap: nope, there still the same issue.

You have all LIBGL settings in the Prefs:Env-Archive still.

As well as file "Install_LIB_EGL_to_SYS" have at end of the file that:

copy Prefs/Env-Archive to Sys:Prefs/Env-Archive

So there 2 errors:

1). all LIBGL files should be now in the Prefs/Env-Archive/EGL_wrap directory

2) there should be no need for that second copy command, enough to have previous one, which will copy Prefs/Env-Archive/EGL_Wrap to Sys:Prefs/Env-Archive/EGL_Wrap, just all the LIBGL files should be placed there, and not in the Sys:Prefs/Env-Archive/ , as it then still copies as before

I double checked, it still copy those LIBGL files to the wrong place in the root of env-archive.

Maybe you just upload old version, and not new one ?

@kas1e

Rhooooo!!!! excuse me the problem is just on installer.
Now i have fixed an envarc internal functions.
I create a new archive for you
Please wait 10 minutes
HunoPPC

@kas1e

It's good now

Best regards

HunoPPC

@Huno
Yeah, now all fine , thanks :)

Also checked your new reboorn_sp : yeah, new env works fine and i was able to catch necessary shaders with it, thanks !

ps. And Mattew says he will send you privately links on beta version of Nova now