Re: [Jikesrvm-core] [Jikesrvm-regression] 1851 FAILURES

[David P G. <gr...@us...>]

It was 13122 
(http://jikesrvm.svn.sourceforge.net/viewvc/jikesrvm?view=rev&revision=13122) 
that broke ppc64-aix. 13121 runs pre-commit, 13122 fails in the same way 
the night sanity did.

My wild guess would be that by switching to allocating thread stacks from 
PLOS instead of LOS we broke some invariant about stack alignment (8 byte 
instead of 4 perhaps).  Running with more than 1 virtual processor results 
in a seg fault during VM bootup (running with only 1 virtual processor and 
the system works fine).

--dave