On Tue, 28 Jun 2011 12:04:39 -0700, Carl Worth <cworth@cworth.org> wrote:

> I assume by "better" you mean that this one is less fragile to issues
> like the precise formatting of our notmuch.h header file.

Yes, that's what I meant.

> Meanwhile, though, our API *is* declared in the header file, (that's
> what the header file exists for), so I do prefer to base our symbol list
> on it.

Sure, that makes sense to me.  I cherry-picked the sed patch to the
release branch.

> In fact, the ideal would probably be to take the objdump method and use
> it in the test suite to verify that its output matches the contents of
> notmuch.sym generated from the header file.

The second patch being turned into a test sounds like a great idea. I
can do that.

d