Per Monty's code review, it's dubious practice to have a class in which
certain members are sometimes visible, other times not. If these were virtual
methods, or non-static data members, the error would be obvious -- but even
with static data members and non-virtual methods, it looks like an ODR
violation. Extract conditional methods as free functions, as in changeset
07cd70e75473.
Changeset 07cd70e75473 moved LLSD::outstandingCount() and allocationCount() to
free functions so we could turn their visibility on/off via LLSD_DEBUG_INFO.
But on some platforms, without proper LL_COMMON_API declarations visible when
we compile llsd.cpp, those free functions lack proper linkage directives.
Declare LLSD_DEBUG_INFO in llsd.cpp so that when the llcommon library is
built, the free functions get proper linkage -- independent of compilations of
LLSD consumers.
Free functions can be unconditionally compiled into the .o file, but
conditionally hidden in the header file. Static class methods don't have that
flexibility: without a declaration in the header file, you can't compile a
function definition in the .cpp file. That makes it awkward to use the same
llcommon build for production and for unit tests.
Why make the function declarations conditional at all? These are debugging
functions. They break the abstraction, they peek under the covers. Production
code should not use them. Making them conditional on an #ifdef symbol in the
unit-test source file means the compiler would reject any use by production
code. Put differently, it allows us to assert with confidence that only unit
tests do use them.
Put new free functions in (lowercase) llsd namespace so as not to clutter
global namespace.
Tweak the one known consumer (llsd_new_tut.cpp) accordingly.
This establishes that there are no viewer-side unit tests relying on these
methods. The point is to try to clean up the LLSD public API. In the same
vein, remove from LLSD public API a diagnostic method which is nothing more
than an implementation detail for the corresponding LLSD::Impl method. The
same effect can be achieved by making LLSD::Impl a friend of LLSD, moving the
method with the messy signature (classic-C arrays!) into LLSD::Impl itself.
Because new enum values have been added to the LLSD type field, a few external
switch statements must be adjusted to suppress fatal warnings, even though we
never expect to encounter an LLSD instance containing any of the new values.
I made it about a year and a half ago; Zero found it while reading code. I had added a return value to LLSD::insert(), but a) did it wrong, and b) broke the STL-like semantics of insert(). So I've put insert() back to returning void and created LLSD::with(), which does what my earlier insert() did. The compiler then caught all the cases where insert()'s return value were being used, and I changed those to use with() instead.
I added a #ifndef hack to v3math.h to eliminate LLString's from the llhavok project.
I also removed lluuid.h's dependency on LLString (using std::string) instead and
then fixed a bunch of bad dependency fallout on a few files around the project
that suddenly lost their hidden access to some fundamental includes.
The important parts were reviewed with James.
Nat Goodspeed