On Mon, 23 Mar 2026 07:40:46 GMT, Matthias Baesken <[email protected]> wrote:

>> It is new to have a platform-specific JVM API but better that than just 
>> ad-hoc-ly exposing a piece of code in libjvm IMO. If you want to have one 
>> version of this code and you want libawt to obtain it from libjvm then that, 
>> by definition, should be part of the exposed JVM API on AIX.
>
> Hi David, in practise there is quite a bit more exported from libjvm and not 
> only the JVM API (JVM_).
> E.g. the debug helpers, jio_ , AsyncGetCallTrace and some numa methods.
> 
> 
> 000000000088dc30 T AsyncGetCallTrace
> 0000000000787ec0 T blob
> 00000000007890b0 T debug
> 00000000007880b0 T disnm
> 0000000000787f60 T dump_vtable
> 00000000007892a0 T events
> 0000000000789480 T find
> 00000000007899a0 T findbcp
> 0000000000789700 T findclass
> 0000000000789330 T findm
> 0000000000789850 T findmethod
> 00000000007893e0 T findnm
> 00000000007895c0 T findpc
> 0000000000789210 T flush
> 0000000000789d30 T help
> 0000000000b16990 T jio_fprintf
> 0000000000b16a20 T jio_printf
> 0000000000b16430 T jio_snprintf
> 0000000000b16970 T jio_vfprintf
> 0000000000b163f0 T jio_vsnprintf
>  ...
> 0000000000e69120 T numa_error
> 0000000000e69110 T numa_warn

Sure there are other groups of exported symbols - debug helpers and jio_* the 
main ones plus AGCT. But that doesn't mean we should just add another ad-hoc 
entry.

The numa entries are interesting because we need to export those to override 
the libnuma defaults.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/30295#discussion_r2979348398

Reply via email to