🚀 go-pugleaf

>>>>> ">" == Richard Stallman <rms@gnu.ai.mit.edu> writes:

    >>       We are trying to use the output produced/generated from the flow analysis
    >>     phase of the compile to develop a kind of static call tree of our code. 

    >> To generate a call tree as a byproduct of compilation is a tempting
    >> idea, but it has a fundamental flaw.  You can't generate a complete
    >> call graph from any kind of output from cc1 (or cc1plus), because they
    >> don't know about all the calls that are in the source.

    >> For example, if a function contains this:

    >> #ifdef vax
    >>      foo ();
    >> #endif

    >> and you happen to be on a sun, by the time cc1 is running there is no
    >> trace of the fact that a call to foo was present there.

    >> If you rely on incomplete data for doing maintenance, bugs will
    >> result.  When you change the way to call foo, you need to find *all*
    >> the calls, not just those that happen to be enabled on the target you
    >> are using today.

    >> For this reason, I believe cross-referencing ought to be done by
    >> parsing the source directly, *without* going through cpp.

If you don't expand the macros you won't find all the calls either.
Also, I don't know anything about cc1plus internals -- will the flow
analysis phase still contain call_insn's for inlined methods?

Jeff