RFR: 8262298 : G1BarrierSetC2::step_over_gc_barrier fails with assert "bad barrier shape"

[Vladimir Kozlov]

On Tue, 16 Mar 2021 13:01:06 GMT, Nils Eliasson <neliasso at openjdk.org> wrote:

> During eliminate_macro_nodes an allocation and a tightly coupled arraycopy is eliminated. The gc barrier is still there but the entry condition is always false. During the subsequent igvn we can hit an assert depending on what order the barrier nodes are removed. 
> 
> If the entry to the gc barrier is removed first - the exit region node will have Parm-control as in(2). This is a sign that it is a barrier during teardown. I add a check for that in this patch.

Will the barrier be eliminated eventually in this case?

-------------

Marked as reviewed by kvn (Reviewer).

PR: https://git.openjdk.java.net/jdk/pull/3029