Diagnosing reboot under load

[Alex Zbyslaw]

Micah wrote:

> Except it reboot later when rebuilding the KDE stuff with portinstall 
> NOT the Gnome script.  And I'm getting occasional segmentation faults 
> on Thunderbird and intermittent compiler errors like this one while 
> portinstalling kdepim:
> then mv -f ".deps/eudora_xxport.Tpo" ".deps/eudora_xxport.Plo"; else 
> rm -f ".deps/eudora_xxport.Tpo"; exit 1; fi
> eudora_xxport.cpp: In member function `virtual KABC::AddresseeList 
> EudoraXXPort::importContacts(const QString&) const':
> eudora_xxport.cpp:121: internal compiler error: Segmentation fault
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
>
While I agree with the previous poster that the upgrade script is a pig 
(why does it silently update ports which have *nothing* to do with 
gnome?) that is not you problem.  The script may take all day, or all 
week, to finish, but it *should* finish.

You have a *hardware* problem.  Random segfaults are a classic bad 
memory symptom.  Whether you have a memory stick that is going 
permanently, or getting too hot, or not getting enough voltage under 
load I cannot say, but but one of them or something like it is happening.

> Something seems wrong.  I don't think a software error should cause a 
> system reboot without any log messages anywhere. 

With a memory fault (or any other hardware fault) all bets are off.  
Think about it.  If the program and data that you are supposed to be 
running become corrupted, how on earth can the kernel guarantee to write 
something to a log file?  It was doing exactly what it was supposed to 
do and then it got fed garbage.  Garbage in, garbage out - that's been 
true since forever and isn't going to change.

--Alex