RFR: 6879063: SA should use hsdis for disassembly (original) (raw)

yumin.qi at oracle.com yumin.qi at oracle.com
Tue Sep 18 13:37:24 PDT 2012

Thanks. John

In fact there is a warning in build but was ignored, I did not pay attention to it: "hsdis.c", line 127: warning: implicit function declaration: decode_instruction_virtual, this symbol ignored in .c as external so the build passed.

Thanks for the catch. This function is not called, so the flaw will not be found.

I haven't tested the old hsdis.so, thought they work with old JVM. Will have a check. This value currently will never be true, that is right.

Thanks Yumin

On 9/18/2012 12:01 PM, John Rose wrote:

On the whole, I am delighted to see all the code being removed, and pleased to see a fresh purpose for the hsdis plugin.

In hsdis.c, there is a symbol 'decodeinstructionvirtual' (no 's' at the end) which appears to have no definition. Is it a typo? The "nice newlines" have moved around and proliferated. Will this break any compatibility with new JVMs running old hsdis.so or vice versa? I notice that 'donewline' currently does not have a way to be true. Was that meant for the old entry point? But these are minor points. Reviewed. — John

More information about the hotspot-compiler-dev mailing list