example_calltree fix and more resolved symbols

[xavier-martin]

No description provided.

[commial]

Hi,
Thanks for the PR.

It indeed seems there is a weird behavior regarding the CallRetCallback when too much instruction are executed in one run. This PR implements a workaround for this issue.

I'm OK to merge it that way, as it fixes the calltree example.
To sum up a few of the tests made:

• running with a -1 step count is actually using a default stepCount value inside the library
• at the very end, if there is still n instructions before the trace end, and if we asked stepCount > n in replayForward, the callback is not called (but breakpoints are honored, the number of executed instruction is correct, etc.)
  • the fix implemented here is to "discover" how many instruction is missing until the end of the trace, go back, and ask for exactly n stepCount
• if stepCount is too big, after m instructions the callback is no more called. This could have been linked the raise of an event unmasked in the trace or something similar, but using several replayForward with less than m instructions correctly works. The working window is also moved if we start later in the trace.
  • The implemented fix is using an empirically working number for stepCount, which seems to be inside the working window

Also, in the future, the PDB fetching should be moved in the lib as an helper (tracked by #9).

As a result, I'm merging it now. Please, do not hesitate to comment if I'm missing a test case here.