|
[
Permlink
| « Hide
]
David Grove added a comment - 05/May/08 09:27 AM
I'd like to at least understand why this is failing even if we don't fix it before making the 2.9.3 release.
ok. I've poked at it a little bit and I'm not going to be able to clear this up quickly.
Testing old revisions it seems that BaseBasePoisoned worked on ppc32 when poisoning was committed but FullAdaptive did not.
This is on my list to look at this soon, but naturally will not be before 2.9.3. Would like to see some investigation of this before 3.0 if possible (pending confirmation of sanity runs on excalibur over the weekend to see if it really is still a problem).
testing on ppc64-aix. It looks like the problem occurs more or less at the same time as we are making the first JNI calls:
[piccolo:/home/dgrove/buildit/rvm-trunk] ./dist/BaseBasePoisoned_ppc64-aix/rvm -X:verboseBoot=20 The bug manifests as a SIGSEGV (hard crash) on ppc64-linux.
It's happening in the first JNI function that's invoked from the first native method we invoke (OnLoad in GNU Classpath's jcl.c): JNIEXPORT jint JNICALL if ((*vm)->GetEnv (vm, &envp, JNI_VERSION_1_4) != JNI_OK) { return JNI_VERSION_1_4; } env = (JNIEnv *) envp; Poisoned is ok on PPC32-OSX, this issue seems related to AIX so I think Dave will be the only one able to fix it.
I am looking at this, but it also fails on ppc-linux, so it is not AIX specific.
Fixed in r14839.
Leaving open for Dave to double check the change (and confirm it is being included in 3.0) as the release is imminent. (we also need to check the status for 64 bit)
will respin 3.0.0 release branch to pick up fix.
|
||||||||||||||||||||||||||||||||||||||||||||