Julian Wang
Julian Wang
will do, after i return from vacation next Wed.
High-level comments: 1) HelperCallSnippet seemed being used for disjoint purposes across platforms and in OMR side as well. this use (new file) in OpenJ9 PPC codegen looked more aligned with...
still somewhat at a high level: There were J9 helperCallSnippet customization examples already for certain special purposes in files: `runtime/compiler/p/codegen/J9PPCSnippet.[ch]pp` I don't see why you cannot add this case there....
@rmnattas please take a look from JIT perspective
i suspected this fix here: https://github.com/eclipse-openj9/openj9/pull/22370 is applicable to this defect.
could `0x00000000006C9200` be identified to be a J9Class pointer maybe?
now, happy chasing down in the complained JIT-ed method when/why `obj->clazz` is stored into an O-slot.
I think: the first section named as `JIT-RegisterMap` is for the current method in that particular frame, as described by the current GCStackMap's registerMap at that GCPoint. each register's value,...
so, somehow J9Class pointer established into r30 in this method was misunderstood as a java reference (collectable). that is the bug here.
now, this might be relatively straight-forward to pin down, hopefully traceFull of this method didn't change much from run to run. why and how is a J9Class considered collectable. I...