RFR: 8364588: Export the NPE backtracking functionality to general null-checking APIs [v2]

[Chen Liang]

On Tue, 5 Aug 2025 10:07:38 GMT, Johan Sjölen <jsjolen at openjdk.org> wrote:

>> src/hotspot/share/interpreter/bytecodeUtils.hpp line 40:
>> 
>>> 38:   // Slot can be nonnegative to indicate an explicit search for the source of null
>>> 39:   // If slot is negative (default), also search for the action that caused the NPE before
>>> 40:   // deriving the actual slot and source of null by code parsing
>> 
>> Suggestion:
>> 
>>   // Slot can be nonnegative to indicate an explicit search for the source of null.
>>   // If slot is negative (default), also search for the action that caused the NPE before
>>   // deriving the actual slot and source of null by code parsing.
>> 
>> Periods at the end of sentences in comments.
>
> I'd like to see the description for `slot` pushed into an enum, and make any negative number except `-1` explicitly forbidden.
> 
> ```c++
> enum class NPESlot : int {
>   // docs here
>   None = -1,
>   // docs here
>   Explicit // Anything >= 0
> };
> bool get_NPE_message_at(outputStream* ss, Method* method, int bci, NPESlot slot) {
>   if (slot != NPESlot::None) {
>     // Explicit search
>   }
> }

I don't think I know how to use C++ enums. I tried to define this enum in bytecodeUtils.hpp and use it from jvm.cpp, and don't know how in any way I can ever express `BytecodeUtils::NullSlot slot = search_slot >= 0 ? search_slot : ???;` because `NullSlot::Search` or `BytecodeUtils::NullSlot::Search` or `BytecodeUtils::NullSlot.Search` are all erroneous. I don't think this may worth the hassle since the attempt to use C++ enum only creates more pain, unless you tell me how to use it properly.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/26600#discussion_r2254525459