Re: [g-a-devel]at-spi speedups ...

[Bill Haneman <bill haneman sun com>]

On Mon, 2002-12-16 at 15:54, Michael Meeks wrote:
> Hi Bill,
> 
> On Mon, 2002-12-16 at 13:41, Bill Haneman wrote:
> > Thanks for the patch; it looks great *except* that I do not think there
> > should be any ATK usage in cspi.  Generally I think reuse is great but
> > in this case it would cause more interpenetration of our layers, which
> > is not a good thing.
> 
> 	That's a great shame. After talking for some considerable time about
> this I still have no idea why a layered design is in this case a good
> thing, or sufficiently useful that we sacrifice every other metric of
> code quality to id.
...
We talked on IRC at some length about this.  In the interest in getting
this patch in, it'd be OK if you move the ATK usage to libspi and 'wrap'
it as we discussed.  That isn't my ideal solution, but I do recognize
that it has some modest reuse and performance advantages and I can't
think of a truly compelling argument against it.

If you want to do that then please go ahead and commit, as long as all
the explicit ATK usage is kept out of cspi's sources.  Otherwise I can
prepare a patch to implement StateSet's 6 methods without ATK, and you
can commit your other fixes.  It's up to you, I certainly don't want to
lose the fixes in your patch nor do I want to commit them in your stead
unless you don't want to be bothered with a rewrite.


> 	I'm using gcc 3 of some sort; however it's more likely to be the fact
> that I use -O2 that shows more thinko type warnings up. -Werror is a
> nightmare for portability.

I am using gcc 3.2-7 FWIW.

> 
> 	There are also rather a number of poisoning bugs that are fixed in that
> patch, pwrt. not resetting global or shared CORBA_Environments that
> would be worth getting in.

Yes, thanks for finding/fixing.

-Bill

> 	Regards,
> 
> 		Michael.
-- 
Bill Haneman <bill haneman sun com>