Segment fault when using wireless (adhoc mode)

[陈杰 <chenj lemote com>]

Hi, all
I've tried NM-0.7.1-rc1 and NM-0.7.1-rc2, and find that using wireless (adhoc mode) makes NM easily be terminated by signal 11.

Following the log , I've seen two or more "Stage 5 of 5 ... commit" sections (which were outputed by the function <nm_device_activate_stage5_ip_config_commit>) were ouputed one by one. In the last "Stage 5 of 5 ... commit" section, NM said it recived signal 11.

Adding more log output code to the source of NM, I found the following:
1) In the first call of <nm_device_activate_stage5_ip_config_commit>, the ref_count of the instance ip4_config increased by 2 (by the func <nm_device_set_ip4_config>), but decreased by 1 (nm-device.c, line 1438, "g_object_unref (ip4_config)")

2) In the following sequence calls of <nm_device_activate_stage5_ip_config_commit>, the ref_count of the *same* ip4_config decreased by 1 each call.

3) At last, the ref_count of the instance ip4_config reached zero (then ip4_config was freed), but a following call of <nm_device_activate_stage5_ip_config_commit> accessed this ip4_config, then NM was terminated by SIGSEGV.