Re: mpolacek/ebl_dynamic_tag_name

[Roland McGrath <roland at hack dot frob dot com>]

> This looks fine to me. But note that libebl isn't a stable interface. So
> I wouldn't spend too much time adding specific tests for it, unless
> higher level functions really depend on it. So in this case, why not
> (also) actually read the dynamic segment of the testfile you open and
> check the tag names found in there?

That's a good point.  It's arguably anti-useful to have tests like this
that make someone think the ebl interface must not be broken.  The right
way to test this function is just to run readelf.  You can just choose a
test file synthesized to contain all the known names and then run readelf
-d on it to check the output, there is no need to write a unit test like
this.

> Since we are adding more and more tests, it would be good to have a bit
> more structured framework for it. Maybe using autoconf autotest? But I
> don't really have a suggestion/preference for what/which framework would
> be nicest (but please something simple without external dependencies
> except regular build infrastructure tools). Would you be willing to look
> into that?

I have long been intending to convert our tests to autotest.  I just
never got around to it.  I think it would be a good thing to do.
This is an excellent suggestion for a new volunteer like Marek to
look into.  I'm glad you reminded me about it.

It would start with some macros to replace what test-subr.sh
provides, making it simple to write tests.  The hairy thing there is
to do them so that they serve correctly for both 'make check' and
'make installcheck'.  The other nice thing I'd planned to do was
integrate into the test macros some magic that emits the right magic
data for lcov to delineate the coverage data between tests.  I have
some macro magic kicking around somewhere from another old project
for that.


Thanks,
Roland