[jdk21u-dev] RFR: 8330106: C2: VectorInsertNode::make() shouldn't call ConINode::make() directly

[Goetz Lindenmaier]

On Wed, 2 Apr 2025 17:47:47 GMT, Satyen Subramaniam <ssubramaniam at openjdk.org> wrote:

> Backporting JDK-8330106: C2: VectorInsertNode::make() shouldn't call ConINode::make() directly. Adjusts code to call PhaseValues::intcon(), which is required so the node is properly entered in the GVN's hash table and its type is properly recorded in the GVN's type table. Ran GHA Sanity Checks, and local Tier 1, and Tier 2 tests. Patch is clean.

Hi @satyenme 
please be more careful and conservative with your risk ratings.
This change can break compilation on all platforms.  So the risk is much bigger than say https://bugs.openjdk.org/browse/JDK-8331088.  Thus "risk low"  in my eyes is unrealistic. Changes to gc, compiler, interpreter etc in my eyes never have a low risk.  The risk of this change is mitigated by the fact that is is live in 23 without issues which makes it acceptable for backport.

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

PR Comment: https://git.openjdk.org/jdk21u-dev/pull/1579#issuecomment-2775061227