Ocarina of Time - Crash with Glide64mk2 (Bug Tracking)

[Starlord]

Hi everyone,

I updated to the newest build which changed some code in glide64mk2.

+commit 720066252593c91c0dc16b703badcd92dc2a0a49

+Merge: fb7f81d 3a44862

+Author: Dorian Fevrier <fevrier...@yahoo.fr>

+Date:   Sun Dec 21 16:31:12 2014 -0500

This change has a regression with regard to Zelda - Ocerina of Time.
If the player is in Dodongos Cavern and using a bomb near a wall, which can be blown up, mupen64plus and m64py (comment: this frontend is awesome) are crashing. After an downgrade from the glide64mk2 to the lastest revision the issue is gone.

Please find below the log:

Core: Imagetype: .z64 (native)
Core: Rom size: 33554432 bytes (or 32 Mb or 256 Megabits)
Core: Version: 1449
Core: Manufacturer: Nintendo
Core: Country: Europe (0x50)
Frontend: INFO: ScreenSaver disabled
Video: opening ...Glide64mk2.ini

Video: 3DNOW! detected.

Frontend: INFO: using Video plugin: 'Glide64mk2 Video Plugin' v2.0.0
Frontend: INFO: using Audio plugin: 'Mupen64Plus SDL Audio Plugin' v2.0.0
Input: 3 SDL joysticks were found.
Input: N64 Controller #1: Using manual config for SDL joystick 0
Input: N64 Controller #2: Using manual config for SDL joystick 1
Input: No auto-config found for joystick named 'Bluetooth HID Port' in InputAutoConfig.ini
Input: 2 controller(s) found, 2 plugged in and usable in the emulator
Input: Rumble activated on N64 joystick #1
Input: Rumble activated on N64 joystick #2
Input: Couldn't open rumble support for joystick #3
Input: Couldn't open rumble support for joystick #4
Input: Mupen64Plus SDL Input Plugin version 2.0.0 initialized.
Frontend: INFO: using Input plugin: 'Mupen64Plus SDL Input Plugin' v2.0.0
Frontend: INFO: using RSP plugin: 'Hacktarux/Azimer High-Level Emulation RSP Plugin' v2.0.0
Video: Using TEXUMA extension.

Core: Setting video mode: 1280x720
Video: InitCombine()
Video: extensions
Video: initialized.
Video:

Video: Error while changing current directory back to original path of 'C'!
Video: Error while changing current directory back to original path of 'C'!
Audio: Initializing SDL audio subsystem...
Input: Rumble activated on N64 joystick #1
Input: Rumble activated on N64 joystick #2
Input: Couldn't open rumble support for joystick #3
Input: Couldn't open rumble support for joystick #4
Core: Starting R4300 emulator: Dynamic Recompiler
Traceback (most recent call last):
  File "c:\mupen64plus-ui-python.git\dist\windows\build\m64py\out00-PYZ.pyz\m64py.frontend.worker", line 355, in run
  File "c:\mupen64plus-ui-python.git\dist\windows\build\m64py\out00-PYZ.pyz\m64py.core.core", line 346, in execute
WindowsError: exception: access violation reading 0x55882266

Another question I have is about the google project. Is the bug tracking still valid?
If you don't mind, I can enter and update some issues. I saw that some issues are already fixed.

Thanks a lot and best regards,
Starlord

[Dorian FEVRIER]

Thanks Starlod! :)

The code you are pointing only change some documentation:

https://github.com/mupen64plus/mupen64plus-video-glide64mk2/commit/720066252593c91c0dc16b703badcd92dc2a0a49

Can you share at which commit the problem appear please? :) I suspect it's not related to video. Can you repro the bug with --gfx 'dummy' ? same for audio and input.

I'm also VERY interested by any savestate to share. This ways I can check if the problem also appear with Rice (and put a regression test for it). :)

Regarding the google project issue: "Is the bug tracking still valid?"

From my perspective: Yes as it focus on "user" problems and game compatibility. Github mainly focus on "dev" problems...

But:

- mupen64plus is a big (and sometime complicate and old) code set.

- mupen64plus lack of devs understanding this code and try to improve it (make it more maintainable).

- The few time available devs give to the project is very limited and it's easier for new devs to adding features than trying to fix dirty things from the ground.

And all those points are "dev" problems.

I see the google project page and sometime close some issues. I also use it to see what users want. But putting users in Github dev and code discussions is a bad idea IMHO.

Once again, save states are the only way for us to potentially reproduce a problem so please provide them in the google project issue page.

 

--
You received this message because you are subscribed to the Google Groups "mupen64plus" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mupen64plus...@googlegroups.com.
To post to this group, send email to mupen...@googlegroups.com.
Visit this group at http://groups.google.com/group/mupen64plus.
For more options, visit https://groups.google.com/d/optout.

[Starlord]

I have to thank you, for the incredible work. :-)

I tried already with Rice to reproduce the issue. But Rice is working without any issue.
The plugin version of glide64mk2, which reproduces the issue, is from this commit:
https://bitbucket.org/ecsv/mupen64plus-mxe-daily/commits/f051c74a57b6caf067f90dddb6f4fbded30b5c1e

After downgrading to an earlier glide64mk2 version, the issue is gone.
https://bitbucket.org/ecsv/mupen64plus-mxe-daily/commits/878de83773231a652867524cbbd34932a64a51f2

Please find attached a savefile. You are already in Dodongos Cavern.
You have to go/jump to the right part in the entrance hall.
There you can grab a bomb and try to blow the wall away.

My first thought was also it's not related to the video plugin, but the issue can only be reproduced with glide64mk2
(which is really strange, if you look at the log file).

Coming back to Google Bug Tracking.
Yeah, I see. The lack of devs is always a big issue. Also if you find new devs, they have to understand the code first. Especially if it's a long-term project like mupen64plus, this can be really challenging. :)
Then I will mainly enter bugs in googles bug tracker.
I fully agree with you, that entering user issues in a dev tracker will cause more problems.

Thank you,
Starlord

[Dorian FEVRIER]

Thanks! The fact you've been able to identify the commit really help!

I've created a ticket, I guess it's related to a new config type value getted in the config. https://github.com/mupen64plus/mupen64plus-video-glide64mk2/issues/30

Questions:

Can you open your mupen64plus.cfg and copy paste the lines for parameters:

"force_polygon_offset"
"polygon_offset_factor"
"polygon_offset_units"

The problem is maybe here.

Also, please provide the log. :)

[Dorian FEVRIER]

Also, if you remove your mupen64plus.cfg file and reproduce the bug. Does it appear?

If not, close mupen64plus (the mupen64plus.cfg file is recreated at mupen64plus exit) and restart it and retry to reproduce the bug. Does it crash?

[Richard Goedeken]

On Mon, December 22, 2014 1:02 pm, Starlord wrote:
>
> Another question I have is about the google project. Is the bug tracking
> still valid? If you don't mind, I can enter and update some issues. I saw
> that some issues are already fixed.
>
> Thanks a lot and best regards,
> Starlord

Hi Starlord,

The issue tracker on our google code page is still valid, so I would
appreciate any updates that you make.

Richard

[Starlord]

Hi everyone,

so I checked the values in the mupen64plus.cfg.

force_polygon_offset = false
polygon_offset_factor = 0
polygon_offset_units = 0

Creating a new mupen64plus cfg file doesn't solve the issue. The crash is still occurring.
The log is not changing.
I also made a video to see what exactly happens (during the video I am loading a savestate to jump directly to the wall):
https://www.youtube.com/watch?v=E18E3Z9PZsM&feature=youtu.be

Shall I change the values in the cfg file?
I don't know what exactly the new values are doing, that's why I kept them unchanged.

Kind Regards,
Starlord

[Dorian FEVRIER]

Thanks for the video. Unfortunately, this is what I did and can't reproduce the problem... :(

I can only suggest to create a ticket in the google issue tracker. :(

[Starlord]

Beside creating an issue at google, I have another idea.
I can create a package, which contains my files, configurations, libraries etc.
So at least, you can try to reproduce the issue on your PC.

If you cannot reproduce the issue, it's most likely related to my system.
But at least we know, what could be the potential root cause.
Maybe it's a configuration which we didn't thought about.

[Starlord]

Hi everyone,

I created the mupen64plus package.
https://drive.google.com/open?id=0B8IAFx1H_E5PQ0pYTGdLQUl5ekE&authuser=0

I added a readme and screenshots concerning my setup.
Maybe you can reproduce the bug with this version of Mupen64Plus and M64Py.

Best Regards,
Starlord

[Dorian Fevrier]

I like the idea but I'm on linux. I have no Windows system installed and
even if so, I would have troubles to run a debugger on Windows. :(

If at any moment you are able to reproduce the bug on Linux I would
really appreciate to have you packages. :)

Dorian

On 12/24/2014 04:10 PM, konrad....@googlemail.com wrote:
> Beside creating an issue at google, I have another idea.

> I can create a package with my settings, files, libraries...
> At least you can try, to reproduce the issue on your pc with my M64P
> version and configuration.
>
> If you cannot reproduce the issue on your pc, than it could be system
> related.
> Maybe it is a configuration which we didn't thought about.

>
>
>
> Am Mittwoch, 24. Dezember 2014 16:50:50 UTC+1 schrieb Narann:
>

> Thanks for the video. Unfortunately, this is what I did and can't
> reproduce the problem... :(
>
> I can only suggest to create a ticket in the google issue tracker. :(
>
>
> Le Mercredi 24 décembre 2014 5h06, Starlord

> <starlor...@googlemail.com <javascript:>> a écrit :

> <https://github.com/mupen64plus/mupen64plus-video-glide64mk2/issues/30>

>
> Questions:
>
> Can you open your mupen64plus.cfg and copy paste the lines for
> parameters:
>
> "force_polygon_offset"
> "polygon_offset_factor"
> "polygon_offset_units"
>
> The problem is maybe here.
>
> Also, please provide the log. :)
>
> --
> You received this message because you are subscribed to the Google
> Groups "mupen64plus" group.
> To unsubscribe from this group and stop receiving emails from it,

> send an email to mupen64plus...@googlegroups.com <javascript:>.

> To post to this group, send email to mupen...@googlegroups.com

> <javascript:>.

> Visit this group at http://groups.google.com/group/mupen64plus

> <http://groups.google.com/group/mupen64plus>.

> For more options, visit https://groups.google.com/d/optout

> <https://groups.google.com/d/optout>.

>
>
> --
> You received this message because you are subscribed to the Google
> Groups "mupen64plus" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to mupen64plus...@googlegroups.com

> <mailto:mupen64plus...@googlegroups.com>.

> To post to this group, send email to mupen...@googlegroups.com

> <mailto:mupen...@googlegroups.com>.

[Starlord]

Unfortunately I cannot compile a current build for linux. :/
I tried today to find updated ubuntu binaries with the newest changes, but I wasn't successful.

Nevertheless I tried to reproduce the issue on different windows machines.
I can confirm now, that this issue is hardware independent. I was able to reproduce the issues on different PCs.
It seems mupen64plus, m64py or gilde64mk2 are trying to use a memory sector, which is dedicated for windows.
Most likely you cannot reproduce the issue on linux or any other os due to a different memory allocation.

Also the newest builds show the same issue on windows.
I guess the only way is to debug in windows.

[con...@web.de]

I would guess that it is possible to reproduce when compiling glide64mk2 with:

CFLAGS="-g3 -fsanitize=undefined -fsanitize=address" make -C projects/unix all

or

CFLAGS="-g3 -fsanitize=undefined -fsanitize=address" make -C projects/unix OPTFLAGS="" all

See also following bug for a similar problem: https://code.google.com/p/mupen64plus/issues/detail?id=579#c14

[con...@web.de]

On Tuesday, December 23, 2014 10:14:25 AM UTC+1, Starlord wrote:

I have to thank you, for the incredible work. :-)

I tried already with Rice to reproduce the issue. But Rice is working without any issue.
The plugin version of glide64mk2, which reproduces the issue, is from this commit:
https://bitbucket.org/ecsv/mupen64plus-mxe-daily/commits/f051c74a57b6caf067f90dddb6f4fbded30b5c1e

After downgrading to an earlier glide64mk2 version, the issue is gone.
https://bitbucket.org/ecsv/mupen64plus-mxe-daily/commits/878de83773231a652867524cbbd34932a64a51f2

Can you please narrow it down a little bit more? Just go to "download"  "tags" and check the builds between "2014-12-18" (doesn't work according to you) and "2014-12-08" (seems to work for you).

[littleguy77]

On Sunday, January 4, 2015 3:26:43 PM UTC-5, con...@web.de wrote:

Can you please narrow it down a little bit more? Just go to "download"  "tags" and check the builds between "2014-12-18" (doesn't work according to you) and "2014-12-08" (seems to work for you).

This.  There are several commits to different plugins between the two tags you list.  It might be crashing in glide, but it could be related to changes in core, or ui-console, not just glide.  For any eagle-eyed devs out there, here are all the commits between Dec 8 and Dec 18

From core:
https://github.com/mupen64plus/mupen64plus-core/commit/cd8df80a8a7297102673b56231e9210d7adef57c
https://github.com/mupen64plus/mupen64plus-core/commit/777890adffad4b0b463a7f7957a2dcc5a41074bb
https://github.com/mupen64plus/mupen64plus-core/commit/e58b813f5827c193c1bf5734e5029cf0b41c4b61
https://github.com/mupen64plus/mupen64plus-core/commit/c2320e78248368f9f4def5e6fa463d13d8650516
https://github.com/mupen64plus/mupen64plus-core/commit/ec59c1813291b0910fc1caaca3f459746b61ea1c

From video-glide64mk2:
https://github.com/mupen64plus/mupen64plus-video-glide64mk2/commit/fb7f81dcccb0061b8951253e59d1da1344a39645

From ui-console:
https://github.com/mupen64plus/mupen64plus-ui-console/commit/36c51b5c8fed71e67896627b5261add9c1b4411f
https://github.com/mupen64plus/mupen64plus-ui-console/commit/90fefcab64308c7288e9e179a520548643729e1b

[Conchúr Navid]

I've narrowed it down myself and showed that it is easy to have a crash like yours under linux too: https://github.com/mupen64plus/mupen64plus-video-glide64mk2/issues/30#issuecomment-68653779

[Starlord]

Thanks a lot for the support, Conchúr Navid.
I dunno if the newest build contains the code workaround, you suggested in github.
But I have another interesting finding.

Meanwhile I am at Goron City in the future.
Here you can see different memory issues with glide64mk2.
First, if you press start and save, the text in the save screen is garbage.
Second, bombing the ways free is crashing the emulator (as already known).

Again, if I go back to the older version of glide64mk2, both issues are gone.
I will upload a save state as soon as possible. :-)

[Conchúr Navid]

On Monday, January 5, 2015 at 9:28:13 PM UTC+1, Starlord wrote:

Thanks a lot for the support, Conchúr Navid.
I dunno if the newest build contains the code workaround, you suggested in github.

 I was not proposing a workaround on github. This was just a test to see if this is the problem. And no, it is not part of any release. But the problem was already there with the old version but is just not triggered in the old build. Also glide64 is affected and not only glide64mk2. It is just sometimes hidden because the memory/stack layout by the compiler differs slightly.

And no, no one is working on the problem right now. The only way to work around is to check the ext value in the TexLoad*.h files everytime it is calculated and it may have ended up as negative. Doing it for this special tested function *4b is not enough.

But there is a buffer overflow in the shared path detection code and no one is interested in it. So I doubt there is interest in the other crashes

[littleguy77]

There is definitely interest.  This issue is being tracked on github and there is a lot of discussion on the issue.

https://github.com/mupen64plus/mupen64plus-video-glide64mk2/issues/30

It's just that a lot of devs are currently in the thick of big refactorings, dynarec fixes, and syncs with downstream code.  Hopefully when the dust settles someone will have the time to create the appropriate fix.  I certainly appreciate the level of effort Navid has already put into this -- the problem is pinpointed and the "right" fix just needs to be resolved.

[Starlord]

Hi everyone,

I don't know exactly what changed, but with this release the issue is gone.
Thanks for all your work!

Best Regards,
Starlord

[Dorian FEVRIER]

Thanks for report Starlord.

I'm not sure the actual problem is fixed. It's a tricky memory problem that can appear or desepear depending on code modification.

Dorian

--
You received this message because you are subscribed to the Google Groups "mupen64plus" group.

To unsubscribe from this group and stop receiving emails from it, send an email to mupen64plus...@googlegroups.com.
To post to this group, send email to mupen...@googlegroups.com.