Description of problem: Running this script: ==================================== # cat mod.stp probe module("ext3").function("*") { printf("%s\\n",probefunc()) } ==================================== aborts in pass 2: # stap -v mod.stp Pass 1: parsed user script and 44 library script(s) in 410usr/10sys/1174real ms. terminate called after throwing an instance of 'std::logic_error' what(): basic_string::_S_construct NULL not valid Aborted Version-Release number of selected component (if applicable): systemtap-0.7.2-2 kernel-2.6.18-92 (NOT occuring on 2.6.18-124) How reproducible: always Steps to Reproduce: 1. stap -p2 -v mod.stp Actual results: Pass 1: parsed user script and 44 library script(s) in 410usr/10sys/1174real ms. terminate called after throwing an instance of 'std::logic_error' what(): basic_string::_S_construct NULL not valid Aborted Expected results: compiles & runs flawlessly
Created attachment 325560 [details] backtrace backtrace
From the backtrace, we've determined that the problem is that dwarf_diename() returned NULL, which was then passed to a std::string constructor. Unfortunately, NULL isn't a valid value to pass to the constructor. I've audited all uses of dwarf_diename() in systemtap and ensured that a return value of NULL is properly handled. These changes have been checked into the mainline of systemtap development (git commit id 50e22e6). With these changes, mainline systemtap works correctly on s390x kernel-2.6.18-92 (and 2.6.18-124).
The correct git commit id is 0e68eaa.
0.9.* releases already in RHEL5.4 include a fix for this problem.