Minor endian and debug enhancement

[yadutaf]

• add DEBUG_* constants in place of the hardcoded values. Should help making self-documenting code
• move endian fixes to clang, previous PR was merged... fast :)

I was trying to automatically insert bpf_probe_read in situations like https://github.com/iovisor/bcc/blob/master/tools/solisten.py#L105 but did not find the right way. I tried tweaking https://github.com/iovisor/bcc/blob/master/src/cc/frontends/clang/b_frontend_action.cc with no success. Do you have a guide on how to run it into a debugger ?

[4ast]

lgtm

[drzaeus77]

I generally used llvm::errs() to debug, since there is a lot of tree-walking going on such that following in gdb can get confusing. It was easier for me to get a dump of the control flow and stare at it after the fact. That said, the easiest way to get gdb into b_frontend_action is probably to instrument or attach to tests/cc/test_static, since that will be straightforward to test and run.

The probe stuff happens mostly in the ProbeVisitor class (which you may have already come across). Try for instance adding E->dump() wherever memb_visited_.insert(E); occurs. In your case, what I suspect happens is that the call to inet_sk(sk) hides the ProbeChecker::needs_probe_ semantic from the analysis. Otherwise doing evt.lport = inet->inet_sport might have worked automatically.