[jdk17u-dev] RFR: 8361117: SIGSEGV in LShiftLNode::Ideal due to unexpected dead node

[Roland Westrelin]

On Fri, 5 Dec 2025 12:07:37 GMT, Martin Doerr <mdoerr at openjdk.org> wrote:

> This is a new fix for [JDK-8361117](https://bugs.openjdk.org/browse/JDK-8361117) for OpenJDK. I can't see the corresponding fix because it's not public.
> 
> This fix is taken from https://github.com/openjdk/jdk/commit/05a0a710313917fe7124ff43fe9c9af1d649bcac.
> The problem is described in https://github.com/openjdk/jdk/blob/674cc3eeca77f1f2a6d937b1df5c5cd8a13c2d31/test/hotspot/jtreg/compiler/arraycopy/TestACSameSrcDst.java#L87 but that regression test didn't reproduce the issue in jdk17u.
> 
> I've taken the reproducer from the issue and put it into a new regression test. The VM crashes without the fix and the test passes with the fix.

Otherwise, looks good to me.

test/hotspot/jtreg/compiler/c2/Test8361117.java line 37:

> 35: package compiler.c2;
> 36: 
> 37: public class Test8361117 {

Can you rename the test case? The usual guideline is to not name test cases based on the bug id but rather have a short descriptive name.

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

Changes requested by roland (Reviewer).

PR Review: https://git.openjdk.org/jdk17u-dev/pull/4204#pullrequestreview-3552071748
PR Review Comment: https://git.openjdk.org/jdk17u-dev/pull/4204#discussion_r2598624072