On Thu, Jan 21, 2016 at 8:32 PM, Alex Potapenko opotapenko@gmail.com wrote:
On Thu, Jan 21, 2016 at 4:20 AM, Waldemar Brodkorb wbx@uclibc-ng.org wrote: Have you seen that Alpine Linux is carrying a patch with modifies exactly our problematic function: https://github.com/alpinelinux/aports/blob/master/community/openjdk8/icedtea...
Great! Lots of thanks, it's working now! This patch is must-do, strange it hasn't been accepted upstream.
I'm absolutely sure that initLocalIfs() must have minimal fool-proof.
If you want to try to push this patch into upstream, you should (IMHO):
* Ask Timo or Natanael (Alpine Linux maintainers) was the any response from OpenJDK maintainers?
* Try to collect additional information about source of condition in which initLocalIfs() called twice. To do that I suggest: - put breakpoint at initLocalIfs - run program - at break take an backtrace, continue
I can't see, how uClibc might be involved into problem for now. JNI_Onload() should be called from ClassLoader, not by ld constructor...
regards, Leonid