Not working with Chrome 22/Linux #3

mrdoob opened this Issue Sep 28, 2012 · 7 comments


None yet

2 participants

mrdoob commented Sep 28, 2012

I'm loading Super Mario 64.z64 and I'm getting this as soon as it starts executing it:

Error! Exception :Unknown op: 0xa4000430, 0x0020fffd

It used to work with Chrome 21.


Interesting - thanks for the report.

When I step through from boot (with 22.0.1229.79), I see this:

a4000430: 1420fffd BNE at != r0 --> a4000428

The second u32 there is the opcode. It should be 0x1420fffd - it looks like the top byte is getting lost somewhere along the way.

I did report a very similar bug against v8 in Chrome 18:

I wonder if it's a similar bug? What I don't understand is why I'm not seeing the same here with Chrome 22/OSX 10.8.1.

What platform are you running Chrome on?

@hulkholden hulkholden was assigned Sep 29, 2012
mrdoob commented Sep 29, 2012

I'm on Linux.

I used to see similar issues weeks ago on Chrome OS, which was running Chrome 22 dev then. I though it had something to do with Chrome OS, but now, on Chrome 23 dev it works fine. So I guess it'll get fixed in the next stable.

If no one else has reported the issue I bet it's just a Linux problem.


I have a Linux machine here, so when I get a chance I'll see if I can repro it myself. If I can, I'll see if I can make a minimal repro. I can submit that as a new bug report and use it as a little self-test and warn users if they're running in a bugged browser.

(of course, there's a good chance there's a bug in my code somewhere too!)


Hey, I've just tested this with Ubuntu 12.04 (32 bit) with Chrome 22.0.1229.79 and it seemed ok. Can I check what version of Linux you're using? I'm wondering if 32bit/64bit makes a difference (e.g V8 is generating different code?)

mrdoob commented Oct 4, 2012

Yep. It's 64bit. And yes, as far as I know, V8 could generate different code on 64bit.


Thanks, I'll grab the 64 bit .iso and give it a try.

I actually ended up testing by running Ubuntu under Parallels. So I had an exciting russian doll scenario of emulators/virtualisation: MIPS/N64 -> Javascript/Chrome -> x86/Ubuntu -> x64/OSX. Feels like I need another layer in there somewhere.


Just to let you know I've managed to repro this in 64-bit Ubuntu. This is my minimal repro:

It's actually pretty similar to the bug I raised before, but that was related to writes to Uint8Arrays not working correctly, but this seems to be related to reads from Uint8Arrays.

Like you say, it does seem to have been fixed in the latest dev release, so I guess a fix will trickle through in time. I'm just figuring out if it's worth raising anyway, in case it's something they think is worth hotfixing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment