[crac] RFR: 8372933: [CRaC] Restore failed due to incompatible or missing CPU features, try using -XX:CPUFeatures=0xfoo, 0xbar on checkpoint [v5]

[Timofei Pushkin]

On Mon, 2 Feb 2026 15:02:07 GMT, Jan Kratochvil <jkratochvil at openjdk.org> wrote:

>> src/hotspot/share/runtime/crac_engine.cpp line 514:
>> 
>>> 512:         image_features &= *datap;
>>> 513:       }
>>> 514:       log_error(crac)("Restore failed due to incompatible or missing CPU features, try using -XX:CPUFeatures=%s on checkpoint.", image_features.print_numbers());
>> 
>> In the exact case with this wording I believe `datap` should be printed, not `image_features`. Maybe it's a good idea to rename `datap` to `current_features` to avoid confusion, and also make it a reference instead of a pointer.
>
> Fixed the exact case, thanks. Renamed it to `current_features` everywhere.
> 
> I normally used references for any non-null pointer everywhere but OpenJDK does not do that. I haven't found a coding style rule for it but apparently OpenJDK discourages references. Why do you suggest to use a reference here?

I usually use a pointer-to-class/struct parameter only if it either nullable, the function actually needs a pointer (e.g. does pointer arithmetic), or the function mutates the object (i.e. it's a non-const reference). And even the last two are debatable. Otherwise using a pointer makes the syntax bulkier and makes the function declaration less self-documenting (a pointer allows null).

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

PR Review Comment: https://git.openjdk.org/crac/pull/278#discussion_r2754961913