[jdk11u-dev] RFR: 8303466: C2: failed: malformed control flow. Limit type made precise with MaxL/MinL

[Martin Balao]

On Tue, 25 Jun 2024 15:21:50 GMT, Martin Balao <mbalao at openjdk.org> wrote:

> Hi,
> 
> I would like to propose a backport of 8303466 [1] to jdk11u. jdk11u can benefit from this fix and having more accurate limit type information.
> 
> The jdk17u patch does not apply cleanly because of the following:
> 
>  * src/hotspot/share/opto/addnode.hpp
>    * 11u does not have 8223347. Manually backported the MaxLNode/MinLNode class.
> 
>  * src/hotspot/share/opto/macro.cpp
>    * The 11u backport of 8275330 does not add n->Opcode() == Op_Opaque4 to the assertion. Manually applied the change.
> 
>  * src/hotspot/share/opto/convertnode.cpp
>    * 11u does not have 8256730 so the context is different. Manually applied the change.
> 
>  * src/hotspot/share/opto/loopTransform.cpp
>    * 11u does not have 8301074 (replace of NULL with nullptr) so the change does not apply cleanly. Manually applied the change.
>    * 11u does not have 8273115, so the context is different.
>    * The 11u backport of 8262017 has "set_subtree_ctrl(inner_result_long);" instead of "set_subtree_ctrl(inner_result_long, false);" so the context is different.
> 
> In addition, the following changes were made:
> 
>  * C++11 lambda functions is_clamp and is_sub_con replaced with static equivalents.
> 
>  * nullptr replaced with NULL
> 
> No regressions observed in hotspot:tier1.
> 
> Thanks,
> Martin.-
> 
> --
> [1] - https://bugs.openjdk.org/browse/JDK-8303466

The GHA windows-aarch64 failure at build time seems unrelated to this backport. It's a failure in wget that we also observed in Fedora 40 for some URLs.

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

PR Comment: https://git.openjdk.org/jdk11u-dev/pull/2814#issuecomment-2192052103