Charm++ parallel programming system

[Xuehan Xu <xxhdx1985126 AT gmail.com>]

Dear Sirs:
     I modified some BigNetSim's code to enable the hardware multicast.

     I used the bgTrace in the directory you mentioned last time.

     I run the program like this:   "../tmp/charmrun +p2 ../tmp/pgm 1 0 ++remote-shell ssh"

    on node 0, in "state_BGproc::buildTree", it call "CollectiveManager::buildTree(tmsg)"
    on node 1, CollectiveManager::buildTree(tmsg) is executed.

     though the parameter "tmsg->ginfo.group" on node 1 has the same address as on node 0 when i use debug to check it, but it can't access the contents of

     "tmsg->ginfo.group",such as  "tmsg->ginfo.group[0].nodeID" .

     so I wonder if the "class vector" is delivered with its contents just like what it is on node 0, so when the program on node 1 access the pointer in tmsg, it is actually accessing an invalid address, which lead to the segmentation fault.