This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Trivial clean-up patch for stl_tree
Gawain Bolton wrote:
trivial, indeed, but unfortunately doesn't pass make-check-abi,
because the functions are exported...
Well, I guess this patch could be made non-ABI affecting by only
changing stl_tree.h to remove the function declarations... The
changes to tree.cc could be done when ABI changes are allowed.
But like that we are not winning big...
Ok, we can exchange ideas - my aim is to factorize code to reduce
duplication and help save some time compiling by having code moved out
of header files where possible. What are your objectives?
As you may have noticed, at some point, very close to a release, we had
a wrong code bug due to an incorrect, very hacky cast in the SGI code. I
fixed it quickly introducing duplication in the various insert with hint
taking iterator and const_iterator. That should be be definitely
improved. Also, when I fixed 22102 (i.e., implemented DR 233) on top of
that suboptimal infrastructure, probably I made the situation slightly
worse.
As I said, you are welcome to work on the above or anything else in this
area, but for now please make sure the "ABI" in the wide sense (I mean
interoperability between object modules too, not only vs the .so/.a
library) is respected.
Thanks,
Paolo.