Translator test coverage

[William Cohen <wcohen@redhat.com>]

It has been a while since I have run test coverage tests, so I decided 
that it would probably be a good to generate that information again to 
see what parts of the translator are exercised by the translator and the 
runtime tests. After fixing things for the string keyword in the runtime 
tests, I generated the new coverage information. It is posted on 
http://people.redhat.com/wcohen

http://people.redhat.com/wcohen/stap-trans-tests-20060517/
http://people.redhat.com/wcohen/stap-run-tests-20060517/

Some comments about the what I saw in the test results. There are a 
number of options in stap that are unexercised. There should be tests 
that exercise the options:

"-vv" verbose >= 2, this should also trigger a lot additional code coverage

"-p0" and "-p6" a .ko test to check that error checking for pass number

"-I", "-R", "-m", "-r", "-k", "-s", "-c", "-x", "-h"

There isn't much exercise of the throw/catch paths in the code. Either 
parse or semantic errors. Need some additional .ko tests to exercise 
this paths.

The tapsets.cxx and staptree.cxx are the files that have the least 
amount of coverage.

A number of members in mark_derived_probe, mark_builder, and 
hrtimer_derived_probe in tapsets.cxx are not exercised.

The testing for the printf specification could be more complete in 
staptree.cxx, "%X" and "%n" and justification arguments.

staptree.cxx looks to have code to regenerate the script from the 
internal parse tree. None of that code seems to be exercised.

-Will