On Thu, 23 Apr 2026 07:59:12 GMT, Roland Westrelin <[email protected]> wrote:

>> When doing late inlining, `GraphKit::access_load_at()` may be called
>> to populate an `InlineTypeNode`. When it executes, it, sometimes,
>> causes the `map` used by the `GraphKit` instance to change. When that
>> happens, `CallGenerator::do_late_inline_helper()` continues execution
>> with an outdated `map` which causes issues during parsing.
>> 
>> ---------
>> - [x] I confirm that I make this contribution in accordance with the 
>> [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai).
>
> Roland Westrelin has updated the pull request incrementally with one 
> additional commit since the last revision:
> 
>   extra test

src/hotspot/share/opto/callGenerator.cpp line 706:

> 704:         // GraphKit::access_load_at() may be called from 
> InlineTypeNode::make_from_multi() and it may change the map
> 705:         // that arg_kit uses.
> 706:         map = arg_kit.map();

We also do something similar in Compile::build_start_state(), so does it also 
need to be changed?
If the map changes, which map should we call disconnect_inputs on at line 721?  
Only the new map, the old map, or both?

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

PR Review Comment: 
https://git.openjdk.org/valhalla/pull/2356#discussion_r3134846981

Reply via email to