On Fri, Oct 21, 2005 at 10:45:59AM -0600, Tom Tromey wrote: > >>>>> "Dan" == Apache Harmony Bootstrap JVM <[EMAIL PROTECTED]> writes: > > Dan> Another thought about the '__int64' issue. A friend told > Dan> me a while back that this symbol was typedef'ed in a > Dan> header file called 'stdint.h' on GCC, but this does not > Dan> exist on my Solaris box, and I didn't need it anyway. > Dan> For Linux, I suspect that if you go find this symbol in > Dan> the /usr/include tree, that it will probably appear in > Dan> this file (I _think_ this is the correct name, but I am > Dan> not sure). Perhaps this is the way to solve the problem > Dan> compiling the JNI code? > > Harmony is going to have to ship its own jni.h anyway. This is part > of a complete JVM. FWIW there's already one in Classpath. > > The jni_md.h part is typically architecture, OS, and compiler > dependent. That said, maintaining this file is not a big deal. > The actual code in the x86 linux version of the file in Classpath > amounts to 13 lines.
__int64 is a compiler specific extension, afaik, supported by MSVC++, and I believe gcc defines it also somewhere deep down inside. the portable way to get types of well defined sizes is to use C99's inttypes.h and stdint.h. That's what Kaffe does for jni_md.h, and I really need to get the patch up into GNU classpath before the next release. there is a wrapper autoconf macro for older systems on ac-archive. cheers, dalibor topic > Tom >
