RFR: 8316342: CLHSDB "dumpclass" command produces invalid classes

[Ashutosh Mehra]

On Thu, 28 Sep 2023 22:32:29 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:

>> Please review this change to fix the operands of the bytecodes that operate on fields when dumping a class using SA.
>> 
>> Testing: hotspot_serviceability
>> 
>> I have also verified that `test/hotspot/jtreg/serviceability/sa/ClhsdbDumpclass.javaClhsdbDumpclass.java`, which is in the problem list because of this issue, passes with this change.
>> I have also verified it by dumping the class that has getstatic/putstatic bytecodes and comparing the bytecodes manually with the original classfile.
>
> src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/tools/jcore/ByteCodeRewriter.java line 137:
> 
>> 135:                 case Bytecodes._invokedynamic: {
>> 136:                     int cpci = method.getNativeIntArg(bci + 1);
>> 137:                     cpoolIndex = (short) cpCache.getIndyEntryAt(~cpci).getConstantPoolIndex();
> 
> Is this really suppose to be `~cpci` and not just `cpci`?

That's right. See https://github.com/openjdk/jdk/blob/ecb5e8a03f67c92d7956201de1fa7d07cc6af9cb/src/hotspot/share/prims/jvmtiClassFileReconstituter.cpp#L1053 and https://github.com/openjdk/jdk/blob/ecb5e8a03f67c92d7956201de1fa7d07cc6af9cb/src/hotspot/share/oops/constantPool.hpp#L256

I tend to use `JvmtiClassFileReconstituter` as the reference for the code involved in dumping the classfile, and that makes things easier to follow and implement.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/15973#discussion_r1340737723