JDK-8286800 : Assert in PhaseIdealLoop::dump_real_LCA is too strong
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 11,16,17,18,19,20
  • Priority: P5
  • Status: Resolved
  • Resolution: Fixed
  • Submitted: 2022-05-16
  • Updated: 2023-01-04
  • Resolved: 2022-11-15
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
JDK 11 JDK 17 JDK 20
11.0.19-oracleFixed 17.0.7-oracleFixed 20 b24Fixed
Related Reports
Relates :  
JDK-8298736 - Revisit usages of log10 in compiler code
Relates :  
JDK-8249607 - C2: assert(!had_error) failed: bad dominance
Description
JDK-8249607 added code to PhaseIdealLoop::dump_bad_graph to print the real LCA including an "must always find an LCA" assert that now triggered in some cases. We should investigate why that is and adapt the code accordingly.
Comments
Fix request [17u] I backport this for parity with 17.0.7-oracle. Typical risk of a C2 change. The change is not trivial. Related issues are not needed. We should take it to avoid the bug and stay on par. Clean backport. SAP nightly testing passed.
04-01-2023
A pull request was submitted for review. URL: https://git.openjdk.org/jdk17u-dev/pull/1026 Date: 2023-01-03 14:19:04 +0000
03-01-2023
Changeset: decb1b79 Author: Christian Hagedorn <chagedorn@openjdk.org> Date: 2022-11-15 14:44:26 +0000 URL: https://git.openjdk.org/jdk/commit/decb1b79bc475f024a02135fa3394ff97098e758
15-11-2022
A pull request was submitted for review. URL: https://git.openjdk.org/jdk/pull/11015 Date: 2022-11-07 11:21:36 +0000
07-11-2022
A pull request was submitted for review. URL: https://git.openjdk.org/jdk/pull/10472 Date: 2022-09-28 19:04:07 +0000
30-09-2022
Okay, that's no problem. Then I'll consider adding the improved format of the idom dumps (if that was not touched in the upcoming PR) in a separate RFE.
29-09-2022
Hi Christian - one of our team members is planning to open up a PR for this today.
28-09-2022
Hi [~bstafford], I'm currently working on JDK-8285835 which is a bad graph assert and dumps additional information with `dump_real_LCA`. I also hit this "must always find an LCA" assert. On top of that, I'm not satisfied with the provided information and its format - it's too noisy to make use of it. So, I've started cleaning the method `dump_real_LCA` up. I've now realized that there is already a bug for that assigned to you. If you have not already started with that, could I take this one over?
28-09-2022
ILW = Assert in debug code (that is only executed when asserting due to a bad graph), rare, no workaround = LLH = P5
16-05-2022
See JDK-8278413 and JDK-8283386.
16-05-2022