-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Tom Tromey wrote: > I would assume that the problem is that we're somehow picking up the > wrong jni_md.h -- I would say that we're missing jni_md.h but I would > expect another error message if that were the case. This was it, I think. You of course were able to figure it out faster than I did, but I didn't get your email until later. For whatever reason, jni_md.h in java-gcj-compat was not symlinked to the real jni_md.h. I checked the java-gcj-compat spec and it does appear to be setting the links up correctly, and after reinstalling the java-gcj-compat rpm, it seems fine. Inside jni.h there's an #include <jni_md.h>, and these are in the internal compiler directories. So, if the -I/usr/lib/jvm/java-1.4.2-gcj specified on the command line takes precedence over the internal compiler directories, then the jni_md.h -> jni.h header would have been included instead of the real jni_md.h in the internal compiler directory (without warning I guess...). Now, my larger issue seems to be I am running out of memory during the aot-compile-rpm stage of the eclipse build (I have 1G RAM). The problem is that the machine doesn't necessarily lock up but it crawls for a few hours (very unresponsive to mouse movements or key presses). Is it really normal for the system to become so unresponsive like this? A rogue process like this should eventually be killed by the kernel at least. Eventually, I am getting an internal compiler error, so it might not be available memory after all. I will try with the latest gcc rpm now. - -- Sincerely, David Walluck <david@xxxxxxxx> -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (GNU/Linux) iD8DBQFDvunEarJDwJ6gwowRAulSAJ9yUv5Ar6icCDm6KTZHw1lVrDZu+QCbB6lr /jOAeL9FPHNcnA7yjN9nrN0= =M9DZ -----END PGP SIGNATURE-----
