RFR(S): 8064703: crash running specjvm98's javac following 8060252
goetz.lindenmaier at sap.com
Tue Dec 2 10:50:17 UTC 2014
The change looks good!
Good you do some extra tuning by passing 'true' to maybe_cast_profiled_obj().
I ran some tests on ppc with it, and saw no problems any more.
From: hotspot-compiler-dev [mailto:hotspot-compiler-dev-bounces at openjdk.java.net] On Behalf Of Roland Westrelin
Sent: Montag, 1. Dezember 2014 13:29
To: hotspot-compiler-dev at openjdk.java.net compiler
Subject: RFR(S): 8064703: crash running specjvm98's javac following 8060252
If an arraycopy is used to initialize a just allocated array, when possible we skip the array initialization and let arraycopy do the initialization. If an uncommon trap is triggered after the allocation and before the arraycopy, then an uninitialized array can escape the compiled method to the interpreter. This is fixed by making the uncommon trap resume execution at the allocation rather than the arraycopy and redo the allocation (Vladimir’s suggestion).
More information about the hotspot-compiler-dev