Mol Freeze (cont.)

[Francois Taiani]

Hi Samuel,

>>I was wondering how I could get more info on the context of these 
>>freezes. I can attach a gdb process to the mol session, but the info I 
>>can get from there seems to be quite thin 
> 
> Well, it really isn't MOL that is stuck but MacOS. Thus gdb
> won't tell you anything useful.

That was my guess, so I did not push it very far anyway.

> 
> 
>>I've seen a debugger comes along 
>>with mol, but it looks quite obsure to me.
> 
> The MOL debugger is the best tool [...]
> A few things to check:
> 
> - The kernel version. Older 2.4 kernels have various MM
> problems. MOL exercises the MM layer quite a bit...

I'm using 2.4.18 with the powerpc patch.

> 
> - Make sure you are loading the "MacOS ROM" file 
> from the boot volume. Usage of older MacOS ROM files sometimes
> work but the mismatch will usually cause random freezes.

I've put the line " macos_rompath: 'Dossier Systeme/Mac OS ROM' " (sorry 
for the french bits) in my molrc file, following the advice from the 
console output that told me the boot would be faster then. Maybe I 
should remove it in case that ROM is not good any more. Anyway I'm not 
using the newworld_rom keyword, so I think the ROM is loaded from the 
boot volume, and not from Linux.

> - A not-too-old glibc version. The pthread support used to
> be quite buggy.

I'm using 2.2.5-3 (from debian).

> What hardware do you have btw (in particular, how much memory)?

I'm working on a G3 Blue & White, with 192 Mb RAM. I'm struggling to get 
more memory, since I find it's not that much, but the way things go at 
office I still have some persuasion work left before I actually get it. :)

I'm going to try without any VM, as I did for the first bug, and report 
you what happen. Moreover it seems the freeze happens more particularly 
when I save a file in Word (the file is actually saved, but MacOS 
freezes just after), if I've use the network interface before (though 
that's not totally reproducible).

> 
> Cheers,

Thanks for your hints.

Francois

-- 
Francois Taiani              LAAS-CNRS (http://www.laas.fr)