[LLVMdev] LLVM on OpenBSD

Joachim Durchholz jo at durchholz.org
Thu Jun 19 01:08:26 PDT 2008

Am Donnerstag, den 19.06.2008, 01:13 +1000 schrieb Andrew Dalgleish:
> The build also stops at exactly the same point in several different
> *virtual* machines.
> (the assert() in utils/TableGen/CodeGenDAGPatterns.cpp line 932)

*That* makes a software problem much more probable.

> Please stop repeating the "bad memory" mantra,

Bus errors with almost-random behaviour are typical of marginal
No mantra here, just sticking with the explanation that's the most
probably one matching a specific description.

A VM abort doesn't fully exclude bad hardware.
To convince the last sceptic, you could run the same VM on two different
hardwares (make them as different as possible, i.e. different build
years, different vendors, different models of main board / RAM).
That would exclude the possibility that the VM maps its virtual RAM to
the always same faulty real RAM cell (or some other unknown faulty
behaviour). It would certainly exclude all hardware failures that I
could imagine :-)


More information about the llvm-dev mailing list