But I'm very surprised because the LD_ASSUME_KERNEL=2.2.5 must desactive the nptl option and that work fine in RH 9.0 but not in Fedora Core.
Moreover this option is part of the Fedroa release notes.
The problem with IBM's Java 1.3.1 is no longer limited to NPTL. I saw a bugzilla report that I can no longer find which contained entries from Red Hat developers and IBM Java developers. It indicated that IBM's Java uses some glibc call which used to return a bullshit value (sigaltstack, maybe?). glibc now properly supports the function and returns a proper value. IBM's Java, however, handles the returned value entirely in the wrong manner, so it crashes... regardless of whether you're using NPTL or not.
Wish I could find the Bugzilla report...
