Re: [PATCH] Add ternary search trees to libiberty

[DJ Delorie <dj at redhat dot com>]

What I meant was, have you thought of any other projects (gcc,
binutils, newlib, cygwin) that could use it, and thus is it being
developed in a project-independent way and does it make sense to
burden those project with yet more code they won't use?  I don't mind
adding generic things to libiberty (within reason), as long as they'll
get used by more than one project.  However, things in libiberty are
expected to meet higher standards than things private to a single
project, and if it's only going to be used by one project anyway you
will be adding some overhead (for yourself and for other projects) if
you put it in libiberty.

> In short, there's a lot of places we could use it besides GDB.

Examples?  Aside from stringpool (I assume you mean gcc's); I'd be
interested in seeing benchmarks on that change.  GCC's performance is
a hot topic these days.

> We have expandable hash tables in libiberty, why not TST's?

The hash table code is before my time, although they seem to be used
only by gcc at the moment.  If you search the archive, you'll see that
I'm pushing back on gcc to reconsider gcc-specific things in libiberty
also.  And "but X is in there" isn't a valid reason to add something.
It should have its own justification.