Re: [rfa] libiberty: splay tree performance improvement

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

> [...]
> > There are no references to the type name outside libiberty in
> > either gcc or gdb code bases.  Can you suggest another
> > nonexistence assurance test?
> 
> Hmm... I wonder if we can tag the structure such that it has an
> indeterminate size when used outside of libiberty/splay-tree.c, so
> that attempts to take sizeof(splay_tree_s) fail?

Well, the canonical method is to not have the declarations of these
structs within the public header at all, just to use opaque pointers
within the API, which by the way the code looks quite ready for.

With the GDB nonuse news, is the patch ok to commit (with the struct
field additions moved to the bottom as requested)?


- FChE

Attachment: pgp00000.pgp
Description: PGP signature