[SOLVED] I am having segfaults in hwinfo. I know it worked before.

business_kid · 03-24-2020, 10:50 AM

He hasn't done much for you yet, I gather. Buggy asm code is very easily done and a bitch to find.

I did some Assembler back in the day; it's the sort of stuff that gives you a headache before you start work on it. I was doing machine code for microcontrollers, so I was the complete OS, and the only I/O was toggling CPU legs. There was some great stuff done for CP/M in Assembler.

I'm doing a web search for my issue, and I'll restore the whole disk if I have to. I hope he sorts you out.

hazel · 03-24-2020, 11:49 AM

He says he's fixed it. That was quick work! I need to test out the new version but I don't know exactly how to check it out. I don't understand git at all.

I've asked him for detailed instructions.

He has fixed it! I don't know what he did but he told me how to check out the new version and it works. I just tested it out for him. I'm gobsmacked! I didn't know this kind of thing could get done so quickly.

hazel · 03-24-2020, 01:48 PM

Well, I think I partly understand this now. There's an assembler instruction called cpuid that overwrites something. This was allowed for in the code; the register it overwrites is saved elsewhere and then written back.

Code:

  "push %%rbx\n\t"
  "cpuid\n\t"
  "mov %%ebx,(%%rsi)\n\t"
  "mov %%ecx,4(%%rsi)\n\t"
  "mov %%edx,8(%%rsi)\n\t"
  "pop %%rbx"

But the restore doesn't work correctly for reasons that are obscure to me.

Quote:

As modern optimizing compilers don't bother anymore with adjusting the stack pointer when they allocate local variables - they just 'know' that there is free space - the rbx register gets saved/restored from the same memory the signature[] array is using. So that part never worked.

The bug is apparently an old one but it popped up in my machine because my compiler decided to put hd_data into that register.

business_kid · 03-24-2020, 02:38 PM

Code:

git clone <url>

usually does it from a console or you can surf there and there's a 'clone' button. Surf to github.com, search for your project in the search bar. Very intuitive. Even I can manage it. Occasionally you have to cobble up the makefiles - 'xmkmf -a' or some such. I only find it a pain if they use cmake, which an increasing number are doing.

EDIT: On the Assembler, this is cheeky, but would this do it

Code:

 "push %%rbx\n\t"
  "push %%rbx\n\t"
  "cpuid\n\t"
  "mov %%ebx,(%%rsi)\n\t"
  "mov %%ecx,4(%%rsi)\n\t"
  "mov %%edx,8(%%rsi)\n\t"
  "pop %%rbx"
  "pop %%rbx"

The idea is to use 2 stack spots, only one of which gets overwritten

hazel · 03-24-2020, 03:05 PM

No need for that. It's been fixed now. He sent me the git instructions and I was able to update and test the new branch out. It works. I've sent him the output in case he's interested.

What I find amazing is that a minor bug (it doesn't seem to have affected anyone else) has been cleared so fast all because I complained. Talk about communal effort!

business_kid · 03-25-2020, 08:51 AM

Well, it's good it's fixed. I doubt if M$. or Apple would bother unless their programmer got the bug. Mind you linux users have the logs, the detailed errors, and a lot of dev tools available.

My usb dev mentioned he had been getting complaints about the bug in post #14, had tried all he could, and wouldn't put more time on it until he had some information to work with. We got him that. Building the kernels was the large headache, and time consuming.

But would you as a programmer rest easy if you knew a bug in your code was upsetting folks? I don't think so.

hazel · 03-25-2020, 10:47 AM

The only programs I've written were for my own use. There are a couple of them available for download on my website but they're basically jeux d'esprit and they probably do have bugs in them. We're talking here about an "official" program that's part of the Linux ecosystem. It came from Suse originally.

You're certainly right about Microsoft. You report a bug to them and it just goes into a queue. If you're lucky, it gets dealt with months later in the next service pack.

camorri · 03-25-2020, 11:25 AM

Hazel,

Just wondering if you could post the URL for getting the updated and fixed hwinfo source.

I removed and reinstalled hwinfo from SBo, that still fails. I guess the maintainer for hwinfo needs to be informed of the now fixed source.

Thank-you for your work on this problem. It was well over my head.

hazel · 03-25-2020, 11:32 AM

You can get it via git from https://github.com/openSUSE/hwinfo. git clone and then git checkout sw_07. You'll have to build it with "make static", otherwise the build crashes.

Presumably there will soon be a new release with the fix in it.

camorri · 03-25-2020, 11:35 AM

Thank-you!

vesperto · 01-06-2021, 07:21 PM

Quote:

Originally Posted by hazel

What I find amazing is that a minor bug (it doesn't seem to have affected anyone else) has been cleared so fast all because I complained. Talk about communal effort!

I've been affected. Well, my old laptop to be precise.
I've poked the SBo maintainer.