On 11/03/2009 04:27 AM, Michael Starzinger wrote:
* Are there any plans to install the jvmti.h file (just like it is done with jni.h) or are there circumstances preventing such an installation?
That is probably just an oversight.
* Why does the mentioned header include jvmti_md.h? All machine-dependent definitions should have already been covered by jni_md.h. Should the missing jvmti_md.h file be provided by the VM?
A VM may need to specialize some behavior in jvmti.h. Gcj did this, and it provides the missing jvmti_md.h. If jni_md.h supplies everything necessary, a "blank"/nop jvmti_md.h will suffice.
I also stumbled upon a mailing list post[1] dealing with this topic. But it seems to me that only a small part[2] actually made it into the GNU Classpath source, the rest remained in the GCJ source.
I think this is self-explanatory now: jvmti_md.h is VM-dependent and should be supplied by the VM vendor. That's why that file never made it into classpath. It wouldn't help any other VM.
Keith
